Percentage of home health episodes of care during which the patient improved in ability to ambulate. This is a rate/proportion measure targeted at older adults with multiple chronic conditions during home health quality of care episodes.
- Measure TypeComposite MeasureNoElectronic Clinical Quality Measure (eCQM)Level Of AnalysisCare SettingMeasure Rationale
Many patients who receive home health care are recovering from an injury or illness and may have difficulty walking or moving around safely. They may need help from a person or special equipment (like a walker or cane) to accomplish this activity. Home health care staff can encourage patients to be as independent as possible and can evaluate patients’ needs for, and teach them how to use, special devices or equipment to help increase their ability to perform some activities without the assistance of another person. Safe ambulation and mobility are critical to being able to remain at home. Improving functional status such as a patient’s ability to perform ambulation/locomotion, contributes to quality of life and allows them to live safely and as long as possible in their own environment. Getting better at walking or moving around may be a sign that they are meeting the goals of their care plan or that their health status is improving. Recovering independence in walking or moving around with assistive devices is often a rehabilitative goal for home health patients, making it a reasonable evaluation indicator of effective and high-value home health care.
MAT output not attachedAttachedData dictionary not attachedYesNumeratorNumber of home health episodes of care where the value recorded on the discharge assessment indicates less impairment in ambulation locomotion at discharge than at start (or resumption) of care.
Numerator DetailsThe number of home health episodes of care from the denominator in which the value recorded for the OASIS item M1860 (“Ambulation/Locomotion”) on the discharge assessment is numerically less than the value recorded on the start (or resumption) of care assessment, indicating less impairment at discharge compared to start/resumption of care.
DenominatorNumber of home health episodes of care ending with a discharge from the agency during the reporting period, other than those covered by generic or measure-specific exclusions.
Denominator DetailsHome health episodes ending with a discharge during the reporting period (M0100[2]=09), other than those covered by generic or measure-specific exclusions.
Denominator ExclusionsAll home health episodes for which the patient, at start/resumption of care, was able to ambulate/locomote independently (M1860[1] = 00), or the patient was nonresponsive (M1700[1] = 04 or M1710[1] = NA or M1720[1] = NA), or the episode is covered by the generic exclusions (see following section).
Denominator Exclusions DetailsHome health episodes of care for which (1) at start/resumption of care, OASIS item M1860 "Ambulation/ Locomotion" = 0, indicating that the patient was able to ambulate independently; or (2) at start/resumption of care, OASIS-E item M1700 "Cognitive Functioning" is 4, or M1710 "When Confused" is NA, or M1720 "When Anxious" is NA, indicating the patient is non-responsive; or (3) The patient did not have a discharge assessment because the episode of care ended in transfer to inpatient facility or death at home; or (4) the episode is covered by one or more of the generic exclusions:
- Pediatric home health patients (less than 18 years of age).
- Home health patients receiving maternity care only.
- Home health patients receiving non-skilled care only.
- Home health patients for which neither Medicare nor Medicaid are a payment source.
- The episode of care does not end during the reporting period.
- If the home health agency sample includes fewer than 20 episodes after all other patient-level exclusions are applied, or if the agency has been in operation less than six months, then the data is suppressed from public reporting on Home Health Compare.
- Hospice exclusion: Episodes of care that end in a non-institutional hospice on or after January 1, 2023 are excluded: M2420 “Discharge Disposition” is 3 and M0100 “Reason For Assessment” is 9.
Table 1 (see supplemental attachment page 3) provides the episode counts by exclusion criterion for episodes of care that started and ended in CY 2022. In CY 2022, 1,658,640 episodes of care were excluded from the denominator for Improvement in Ambulation/Locomotion (#0167) due to meeting at least one exclusion criterion. Approximately 94,000 episodes of care in CY 2022 ended in a discharge to non-institutional hospice. This exclusion criterion is only applicable to episodes of care ending on or after January 1, 2023 and is therefore not listed in Table 1 (see supplemental attachment page 3).
Type of ScoreMeasure Score InterpretationBetter quality = Higher scoreCalculation of Measure Score1. Define an episode of care (the unit of analysis): Data from matched pairs of OASIS assessments for each episode of care (start or resumption of care paired with a discharge or transfer to inpatient facility) are used to calculate individual patient outcome measures.
2. Identify target population: All quality episodes of care ending during a specified time interval (usually a period of twelve months), subject to generic and measure-specific exclusions. Cases meeting the target outcome are those where the patient is more independent in ambulation/mobility at discharge than at start/resumption of care: M1860_CRNT_AMBLTN [2] < M1860_CRNT_AMBLTN [1].
3. Aggregate the Data: The observed outcome measure value for each home health agency is calculated as the percentage of cases meeting the target population (denominator) criteria that meet the target outcome (numerator) criteria.
4. Risk Adjustment: The expected probability for a patient is calculated using the following formula:
P(x) = 1/(1 + e-(a+Σbixi))
Where:
P(x) = predicted probability of achieving outcome x
a = constant parameter listed in the model documentation
bi = coefficient for risk factor i in the model documentation
xi = value of risk factor i for this patient
Predicted probabilities for all patients included in the measure denominator are then averaged to derive an expected outcome value for the home health agency. This expected value is then used, together with the observed (unadjusted) outcome value and the expected value for the national population of patients for the same data collection period, to calculate a risk-adjusted outcome value for the home health agency. The formula for the adjusted value of the outcome measure is as follows:
X(Ara) = X(Aobs) + X(Nexp) - X(Aexp)
Where:
X(Ara) = Agency risk-adjusted outcome measure value
X(Aobs) = Agency observed outcome measure value
X(Aexp) = Agency expected outcome measure value
X(Nexp) = National expected outcome measure value
If the result of this calculation is a value greater than 100%, the adjusted value is set to 100%. Similarly, if the result is a negative number the adjusted value is set to zero.
Measure Stratification DetailsThe measure is not stratified.
All information required to stratify the measure resultsOffAll information required to stratify the measure resultsOffTesting Data SourcesData Sourceshttps://www.cms.gov/medicare/quality/home-health/oasis-data-sets
The reporting of quality data by home health agencies (HHAs) is mandated by Section 1895(b)(3)(B)(v)(II) of the Social Security Act (“the Act”). Outcome and Assessment Information Set (OASIS) reporting is mandated in the Medicare regulations at 42 C.F.R.§484.250(a), which requires HHAs to submit OASIS assessments to meet the quality reporting requirements of section 1895(b)(3)(B)(v) of the Act. It is important to note that to calculate quality measures from OASIS data, there must be a complete quality episode, which requires both a Start of Care (initial assessment) or Resumption of Care OASIS assessment and a Transfer or Discharge OASIS assessment.
Minimum Sample SizeNot applicable.
- Evidence of Measure Importance
Many patients who receive home health care are recovering from injury, surgery or an illness that affects their ability to safely ambulate in their home environment and beyond. Decreased ability or difficulty with ambulation can lead to an increased risk of falls, hospitalization, and functional decline which also increases the risk of becoming homebound, particularly in older adults (Leppä et al., 2021; Meijers et al., 2012; Robinson et al., 2022). Difficulty with ambulation is one of the main reasons that patients are referred to post-acute care services like home health (Prvu Bettger et al., 2015). Mobility directly impacts performance of activities of daily living (ADLs) like transferring, toileting, and bathing as well as instrumental activities of daily living (IADLs) like shopping, preparing meals, and doing housework. Ambulation is an essential part of home health patients’ safety, contributes to quality of life and allows them to live as long as possible in their home environment. Interventions offered by home health such as physical and occupational therapy are effective strategies for helping patients maintain or improve their ability to ambulate safely in the home. The utilization of physical therapy services in home health is associated with an increased likelihood of being able to successfully be discharged into the community without hospital readmission (Knox et al., 2022). Measures related to ambulation and ADL function have been shown to be related to the overall quality of a home health agency and can help detect disparities in care for populations using home health services (Chase et al., 2018; Fashaw-Walters et al., 2022).
Chase, J.-A. D., Huang, L., Russell, D., Hanlon, A., O’Connor, M., Robinson, K. M., & Bowles, K. H. (2018). Racial/ethnic disparities in disability outcomes among post-acute home care patients. Journal of aging and health, 30(9), 1406-1426.
Fashaw-Walters, S. A., Rahman, M., Gee, G., Mor, V., White, M., & Thomas, K. S. (2022). Out Of Reach: Inequities In The Use Of High-Quality Home Health Agencies: Study examines inequities in the use of high-quality home health agencies. Health Affairs, 41(2), 247-255.
Knox, S., Downer, B., Haas, A., & Ottenbacher, K. J. (2022). Home health utilization association with discharge to community for people with dementia. Alzheimer's & Dementia: Translational Research & Clinical Interventions, 8(1), e12341.
Leppä, H., Karavirta, L., Rantalainen, T., Rantakokko, M., Siltanen, S., Portegijs, E., & Rantanen, T. (2021). Use of walking modifications, perceived walking difficulty and changes in outdoor mobility among community-dwelling older people during COVID-19 restrictions. Aging Clinical and Experimental Research, 33(10), 2909-2916.
Meijers, J. M. M., Halfens, R., Neyens, J. C., Luiking, Y., Verlaan, G., & Schols, J. (2012). Predicting falls in elderly receiving home care: the role of malnutrition and impaired mobility. The journal of nutrition, health & aging, 16(7), 654-658.
Prvu Bettger, J., McCoy, L., Smith, E. E., Fonarow, G. C., Schwamm, L. H., & Peterson, E. D. (2015). Contemporary trends and predictors of postacute service use and routine discharge home after stroke. Journal of the American Heart Association, 4(2), e001038.
Robinson, T. N., Carmichael, H., Hosokawa, P., Overbey, D. M., Goode, C. M., Barnett Jr, C. C., . . . Jones, T. S. (2022). Decreases in daily ambulation forecast post-surgical re-admission. The American Journal of Surgery, 223(5), 857-862.
Table 1. Performance Scores by DecilePerformance Gap Overall Minimum Decile_1 Decile_2 Decile_3 Decile_4 Decile_5 Decile_6 Decile_7 Decile_8 Decile_9 Decile_10 Maximum Mean Performance Score 0.798 0.000 0.678 0.729 0.753 0.779 0.796 0.824 0.839 0.858 0.863 0.861 1.000 N of Entities 7,628 776 757 770 749 762 765 762 763 762 762 N of Persons / Encounters / Episodes 4,407,460 21,568 37,470 59,605 87,796 133,995 199,740 299,379 464,732 778,829 2,324,346 Meaningfulness to Target PopulationThe public can comment on the home health quality program when a notice of proposed rulemaking is published as well as through the consensus-based entity public commenting. No comments have been received during this time period regarding this measure. The target of this performance-based measure is the Medicare-certified home health agency. Functional status was confirmed as a domain of importance for quality measurement at a recent Technical Expert Panel (TEP).
- Feasibility Assessment
This is a long-standing measure in the Home Health Quality Reporting Program. We have not identified any feasibility issues for this measure. The Outcome and Assessment Information Set (OASIS) items for this measure must be completed as part of the OASIS assessment. Responses to the questions are required for data submission to the CMS system. The OASIS burden estimates are reported through rulemaking.
Feasibility Informed Final MeasureOASIS data collection and submission are a requirement of the Medicare Home Health Conditions of Participation. Functional assessment is conducted as part of usual clinical practice, and information on ambulation status used to calculate this measure is recorded in the relevant OASIS items embedded in the home health agency’s clinical assessment. OASIS data are collected by the home health agency during the care episode and submitted electronically to CMS via the Internet Quality Improvement and Evaluation System (iQIES). No issues regarding availability of data, missing data, timing or frequency of data collection, patient confidentiality or implementation have become apparent since OASIS-E was implemented January 1, 2023. No changes have been made to the measure specifications.
Proprietary InformationNot a proprietary measure and no proprietary components
- Data Used for Testing
While all the data used to report results in this form are derived from the OASIS, the periods used to generate results vary. We restrict descriptive characteristics, reliability, and validity to CY 2022, the most recent calendar year of data currently available, and the results generated for exclusions and risk adjustment use CY 2021 data, the data used during the most recent maintenance reevaluation and risk adjustment update.
Differences in DataNot applicable.
Characteristics of Measured EntitiesTable 4 (see supplemental attachment page 10) identifies the publicly reporting home health agencies by size and Census region. This distribution of home health agencies is used for reliability and validity testing. 7,628 home health agencies have 20 or more quality episodes starting and ending in CY 2022.
Characteristics of Units of the Eligible PopulationTable 5 (see supplemental attachment page 10) identifies the patient characteristics of quality episodes treated by publicly reporting home health agencies. Characteristics are reported by sex, race, age, and Census region. This distribution of quality episodes is used for reliability and validity testing. 4,407,460 quality episodes started and ended in CY 2022 and met the denominator exclusion and public reporting requirements.
-
- Level(s) of Reliability Testing ConductedMethod(s) of Reliability Testing
Below, we address reliability at two levels: (1) the performance measure and (2) the underlying data element: OASIS item M1860 (Ambulation/Locomotion: Current ability to walk safely, once in a standing position, or use a wheelchair, once in a seated position, on a variety of surfaces).
Reliability of the Performance Measure Score: Abt measured the extent to which differences in each quality measure were due to actual differences in agency performance versus variation that arises from measurement error. Statistically, reliability depends on performance variation for a measure across agencies, the random variation in performance for a measure within an agency’s panel of attributed beneficiaries, and the number of beneficiaries attributed to the agency. High reliability for a measure suggests that comparisons of relative performance across agencies are likely to be stable over different performance periods, and that the performance of one agency on the quality measure can confidently be distinguished from another. Potential reliability values range from zero to one, where one (highest possible reliability) means that all variation in the measure’s rates is the result of variation in differences in performance across agencies, while zero (lowest possible reliability) means that all variation is a result of measurement error.
To assess measure reliability, Abt used a split-half reliability test. First, we randomly divided each publicly reporting home health agency’s quality episodes into two separate equally sized groups. Then, we calculated risk-adjusted performance rates for each group. Then, using the paired performance rates, we calculated the absolute agreement intra-class correlation statistic or ICC(2,1) with a Spearman-Brown correction to address the artificial reduction in home health agency size by half. Additionally, we recalculate ICC(2,1) within each agency size decile, where size is measured as the number of quality episodes treated after denominator and public reporting exclusions.
- Reliability of the Underlying Data Element: The measure is calculated by comparing patient functioning at the start and end of a home health quality episode, as reported by the home health OASIS data set. Patient ability to ambulate is based on response to OASIS item M1860 (Ambulation/Locomotion: Current ability to walk safely, once in a standing position, or use a wheelchair, once in a seated position, on a variety of surfaces):
- Able to independently walk on even and uneven surfaces and negotiate stairs with or without railings (i.e., needs no human assistance or assistive device).
- With the use of a one-handed device (e.g. cane, single crutch, hemi-walker), able to independently walk on even and uneven surfaces and negotiate stairs with or without railings.
- Requires use of a two-handed device (e.g., walker or crutches) to walk alone on a level surface and/or requires human supervision or assistance to negotiate stairs or steps or uneven surfaces.
- Able to walk only with the supervision or assistance of another person at all times.
- Chairfast, unable to ambulate but is able to wheel self independently.
- Chairfast, unable to ambulate and is unable to wheel self.
- Bedfast, unable to ambulate or be up in a chair.
In 2016 and 2017, Abt and partners conducted a field test of new and existing OASIS items on 12 home health agencies in four states for 213 home health patients.[1]Home health registered nurses and physical therapists, trained by the study team, collected data during home visits at start of care (SOC) or resumption of care (ROC), and/or at discharge. Follow-up visits were conducted within 24 hours of the initial field test visit, by a different registered nurse or physical therapist to test interrater reliability. M1860 was one of the existing OASIS items that was tested. Interrater reliability was assessed for SOC or ROC and at Discharge with a linear weighted kappa. The number patients for which inter-rater reliability could be tested was 105 at SOC/ROC and 83 at discharge.
The kappa statistic is generally considered to be the “gold standard” statistic associated with item reliability as it factors in the possibility of chance agreement. Kappa values are reported as decimal values between 0.00 (poor) and 1.00 (perfect). These can be interpreted using the following seven categories:[2]
- Poor < 0.10
- Slight = 0.10 to 0.20
- Fair = 0.21 to 0.40
- Moderate = 0.41 to 0.60
- Substantial = 0.61 to 0.80
- Near perfect = 0.81 to 0.99
- Perfect = 1.00
[1] Abt Associates (2018). “OASIS Field Test Summary Report: Outcome and Assessment Information Set (OASIS) Quality Measure Development and Maintenance Project.”
[2] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics, 1977. 33(1):159-174.
Reliability Testing ResultsReliability of the Performance Measure Score: The table below summarizes the distribution of reliability scores for the 7,628 home health agencies exceeding the public reporting threshold of at least 20 eligible quality episodes of care.
Reliability of the Underlying Data Element: The inter-rater reliability (weighted kappa) values for M1860 Ambulation/Locomotion was 0.43 at SOC/ROC and 0.67 at discharge.
Accountable Entity-Level Reliability Testing ResultsAccountable Entity-Level Reliability Testing Results Overall Minimum Decile_1 Decile_2 Decile_3 Decile_4 Decile_5 Decile_6 Decile_7 Decile_8 Decile_9 Decile_10 Maximum Reliability 0.933 0.853 0.903 0.929 0.951 0.955 0.961 0.970 0.968 0.983 0.988 Mean Performance Score 0.798 0.000 0.678 0.729 0.753 0.779 0.796 0.824 0.839 0.858 0.863 0.861 1.000 N of Entities 7,628 776 757 770 749 762 765 762 763 762 762 N of Persons / Encounters / Episodes 4,407,460 21,568 37,470 59,605 87,796 133,995 199,740 299,379 464,732 778,829 2,324,346 Interpretation of Reliability ResultsReliability of the Performance Measure Score: The ICC(2,1) statistics exceed 0.800, even within the decile with the smallest home health agencies, suggesting strong reliability and acceptability for drawing inferences about home health agencies.
Reliability of the Underlying Data Element Based on the weighted kappa statistics the inter-rater reliability indicated moderate agreement at SOC/ROC (0.43) and substantial agreement at discharge (0.67). Given the scale of the response to this item (seven possible responses), we conclude that the item achieves sufficient reliability.
- Level(s) of Validity Testing ConductedType of accountable entity-level validity testing conductedMethod(s) of Validity Testing
Below, we address validity at two levels: (1) the performance measure and (2) the underlying data element - OASIS item M1860 (Ambulation/Locomotion: Current ability to walk safely, once in a standing position, or use a wheelchair, once in a seated position, on a variety of surfaces).
- Validity of the Performance Measure Score: Abt assessed the convergent validity of the measure. Convergent validity refers to the extent to which measures that are designed to assess the same construct are related to each other. To evaluate the convergent validity of the measure, Abt calculated the Spearman rank correlations of the Improvement in Ambulation/Locomotion (#0167) measure with other relevant OASIS-based measures and the fee-for-service (FFS) claims-based measure Discharge to Community (#3477) measure.
The Spearman rank correlation assesses the statistical dependence between the rankings of two variables. In our case, we rank home health agencies according to the Improvement in Ambulation/Locomotion (#0167) measure and other home health agency-level measures. High correlation or association between the Improvement in Ambulation/Locomotion (#0167) measure and other functional measures of improvement would be expected and desired. Low correlation would indicate that the measure may not be valid (is not measuring what we think it is measuring). We only expect a positive correlation with Discharge to Community (#3477), as the population differs by payer (FFS versus FFS, Medicare Advantage, and Medicaid) and the numerator criteria measure different outcomes (successful discharge to community versus improvement in function).
- Validity of the Underlying Data Element: The OASIS item M1860: Ambulation/Locomotion has been used continuously as part of the OASIS since 2001. The behaviorally benchmarked responses were updated and improved based on input from clinicians and technical experts. The OASIS instrument has been published in the Federal Register for comment (both items and measures based off those items) and no objections or suggestions for revision have been noted regarding the response options.
The original OASIS item was originally carefully designed for measuring and ultimately enhancing patient outcomes as part of the National OBQI Demonstration project (1995 – 2000). OASIS items were derived by first specifying a set of patient outcomes considered critical by home care experts (e.g., nurses, physicians, therapists, social workers, administrators) for evaluating the effectiveness of care. These outcomes were chosen from the most important domains of health status addressed by home care providers. OASIS data items were developed, tested in hundreds of agencies, and refined for measuring outcomes to evaluate and enhance the effectiveness of home care. OASIS data items and measurement methods were reviewed by multidisciplinary panels of research methodologists, clinicians, home care managers, and policy analysts. Several tests of validity were conducted for each OASIS item, including Ambulation/Locomotion. Validity testing included:
1) Consensus validity by expert researcher/clinical panels for outcome measurement and risk factor measurement
2) Consensus validity by expert clinical panels for patient assessment and care planning
3) Criterion or convergent/predictive validity for outcome measurement/risk factor measurement
4) Convergent/predictive validity: case mix adjustment for payment
5) Validation by patient assessment and care planning
Descriptions for these validation assessments are taken from the “Volume 4: OASIS Chronicle and Recommendation” OASIS and Outcome-Based Quality Improvement in Home Health Care, November 2001, Center for Health Services Research, University of Colorado Health Sciences Center, Denver, CO.
Validity Testing ResultsValidity of the Performance Measure Score: Table 7 (see supplemental attachment page 15) shows the Spearman rank correlations of the Improvement in Ambulation/Locomotion (#0167) measure with other publicly reported measures of home health quality derived from OASIS assessments and Medicare Fee-for-Service (FFS) claims.
Validity of the Underlying Data Element: As noted above,
- Consensus validity: The item was reviewed by panels of researchers and clinicians and was recommended for measuring patient outcomes relevant to home health care provision and quality measurement, or for risk adjustment of outcome analyses.
- Consensus validity by expert clinical panels for patient assessment and care planning: The item was reviewed by a panel of clinical experts and was recommended for inclusion in a core set of data items for patient assessment and care planning.
- Criterion or convergent/predictive validity for outcome measurement/risk factor measurement: The item was tested empirically for use in conjunction with outcome measures or risk factors predictive of patient outcomes. The item was found to be related to other indicators of health status and patient outcomes in a statistically significant and clinically meaningful way.
- Convergent/predictive validity: Case-mix adjustment for payment: The item was tested and is used in the grouping algorithm that, in part, determines the per-episode payment to home health agencies for care provided under the Medicare home health benefit.
- Validation by patient assessment and care planning: The item has been used by clinicians for patient assessment and care planning in several hundred home health agencies and has been reported by practicing clinicians to be effective and useful for these purposes.
Results of these validation assessments are taken from the “Volume 4: OASIS Chronicle and Recommendation” OASIS and Outcome-Based Quality Improvement in Home Health Care, November 2001, Center for Health Services Research, University of Colorado Health Sciences Center, Denver, CO.
Interpretation of Validity ResultsValidity of the Performance Measure Score: As detailed in Table 7 (see supplemental attachment page 15), the Improvement in Ambulation/Locomotion (#0167) measure displays a statistically significant positive correlation with several publicly reported measures that similarly assess patient functioning and Discharge to Community (#3477), which lends evidence to the measure’s validity. It may be that strong performance on the other OASIS-based measures directly leads to an improvement in ambulation/locomotion. It may also be the case that high quality agencies perform well on both the Improvement in Ambulation/Locomotion (#0167) measure and other OASIS-based measures of patient functioning and communication due to cultural or organization-level factors.
Validity of the Underlying Data Element: Item validity was established based on results of testing described above. In addition, the item was also reviewed as part of the OMB/PRA review process for the most recent OASIS data set revision which allowed for two national comment periods (60 days and 30 days) wherein the face validity of the item was supported by the comments received.
- Methods used to address risk factorsConceptual Model Rationale
A patient’s improvement in ambulation/locomotion is dependent on a variety of factors, including social risk factors, clinical and behavioral risk factors, and access to care. The conceptual model shown in Figure 2 (see supplemental attachment page 17) is inspired by a similar conceptual model proposed by the Committee on Accounting for Socioeconomic Status in Medicare Payment Programs. In this conceptual model, social risk factors influence access to care and clinical and behavioral risk factors, as well as the measure itself. In turn, clinical and behavioral risk factors influence health care and resource use, access to care, and the measure itself. Access to care only influences health care and resource use. Finally, the affects the measure through interventions like skilled nursing, therapy, and care coordination. These interventions may be able to address some social, clinical, or behavioral risk factors, in part if not fully.
Improvement in Ambulation/Locomotion (#0167) attempts to measure a home health agency’s ability to improve patient ambulation/locomotion while the patient is in its care; however, because certain factors are outside of its control, we risk-adjust the measure. Risk adjustment is used to promote incentives for home health agencies to provide the same care to patients regardless of patient characteristics at SOC/ROC.
The risk factors that can be fully addressed should not be included in the risk adjustment model because the home health agency is expected to be responsible for addressing that risk factor. For instance, if all other risk factors are identical, a home health agency is expected to provide two patients with identical quality care regardless of race or ethnicity.
By contrast, a patient who is living alone will have different needs than a patient who lives in a congregate setting. While a home health agency is expected to adapt its care to different living situations, it is not expected to address all the needs for a patient living alone, like having professionals on staff readily available during emergencies. Similarly, a home health agency is not expected to influence the patient’s clinical and behavioral status at SOC/ROC. As a result, relevant clinical and behavioral risk factors to ambulation/locomotion are included in the risk adjustment model.
The risk adjustment methodology used is based on logistic regression analysis which results in a statistical prediction model for each outcome measure. For each patient who is included in the denominator of the outcome measure, the model is used to calculate the predicted probability that the patient will experience the outcome. The predicted probability for a patient is calculated using the following formula:
P(x) = 1/(1 + e-(a+Σbixi))
Where:
P(x) = predicted probability of achieving outcome x
a = constant parameter listed in the model documentation
bi = coefficient for risk factor i in the model documentation
xi = value of risk factor i for this patient
Predicted probabilities for all patients included in the measure denominator are then averaged to derive an expected outcome value for the agency. This expected value is then used, together with the observed (unadjusted) outcome value and the expected value for the national population of patients for the same data collection period, to calculate a risk-adjusted outcome value for the . The formula for the adjusted value of the outcome measure is as follows:
X(Ara) = X(Aobs) + X(Nexp) - X(Aexp)
Where:
X(Ara) = Agency risk-adjusted outcome measure value
X(Aobs) = Agency observed outcome measure value
X(Aexp) = Agency expected outcome measure value
X(Nexp) = National expected outcome measure value
If the result of this calculation is a value greater than 100%, the adjusted value is set to 100%. Similarly, if the result is a negative number the adjusted value is set to zero.
For a more detailed summary of risk adjustment specifications including definitions of the risk factors, please consult the Home Health Quality Reporting Program Risk Adjustment Technical Specifications 2024 (PDF).[1]
[1] https://www.cms.gov/files/document/risk-adjustment-technicalspecifications2024.pdfRisk Factor Characteristics Across Measured EntitiesTable 8 (see supplemental attachment pages 17-20) shows the mean and standard deviation of the observed value for Improvement in Ambulation/Locomotion (#0167) by risk factor in CY 2022.
Risk Adjustment Modeling and/or Stratification ResultsThe risk adjustment model was developed using OASIS national repository data from assessments submitted between January 1, 2021, and December 31, 2021 (~6.2 million quality episodes). The population of 6.2 million quality episodes for calendar year 2021 was split in half such that 3.1 million quality episodes were used as a developmental sample and 3.1 million quality episodes were used as a validation sample. The following process was used to identify unique contributing risk factors to the prediction model:
- Risk factors were identified based on OASIS items that will remain or will be added following the transition to OASIS-E. The statistical properties of the items were examined to specify risk factors (e.g., item responses were grouped when there was low prevalence of certain responses). Team clinicians then reviewed all risk factors for clinical relevance and redefined or updated risk factors as necessary. These risk factors were divided into 31 content focus groups (e.g., functional status, Hierarchical Condition Categories, etc.). Where possible, risk factors were defined such that they flagged mutually exclusive subgroups within each content focus group. When modelling these risk factors, the exclusion category was set to be either the risk factor flag for most independent or the most frequent within each content focus group.
- A logistic regression specification was used to estimate coefficients among the full set of candidate risk factors. Those risk factors that are statistically significant at probability <0.0001 are flagged for further review in Step 3.
- Each risk factor flagged in Step 2 was reviewed to determine which one of the two groups its content focus group resided. Either its content focus group was explicitly tiered by increasing severity or it was not. This classification determined which risk factor covariates were kept and which were dropped from the final risk adjustment specification. For content focus groups that are explicitly tiered by increasing severity, either all risk factors are included within a content focus group or none of them. For example, if response option levels 1 and 2 for M1800 Grooming were statistically significant at a probability of <0.0001 for a particular outcome, then response option level 3 for M1800 Grooming was added to the list even if it was not statistically significant. If none of the risk factors within an explicitly tiered content focus group was statistically significant at <0.0001, the entire content focus group was removed from the model.
- A logistic regression was computed on the list of risk factors kept after Step 3 above.
- Goodness of fit and reliability statistics (McFadden’s R2, C-statistic, and Intra-Class Correlation) were calculated to measure how well the predicted values generated by the prediction model were related to the actual outcomes. Separate bivariate correlations were constructed between the risk factors and the outcomes to confirm the sign and strength of the estimated coefficients in the logistic model.
- The initial model was reviewed by a team of at least three experienced home health clinicians. Each risk factor was reviewed for its clinical plausibility. Clinicians were asked about the direction indicated by the coefficient in the risk adjustment model and how it compares to their perceived bivariate relationship given their experience treating patients in the home. Risk factors that were not clinically plausible were revised or eliminated if revisions were not possible.
- The risk factors that were deemed not clinically plausible were revised or eliminated, and Steps 3, 4, and 5 in this process were repeated. The resulting logistic regression equation was designated as the risk adjustment model for the outcome.
- The risk adjustment model was applied to the validation sample and goodness of fit statistics were computed. The statistics were similar to the goodness of fit statistics computed with the development sample. As additional testing, home health agencies were stratified across several observable characteristics, and the distributions of the risk-adjusted outcomes were checked to confirm that values remained similar across strata.
Using CY 2021 data, the updated risk adjustment model specification yielded a McFadden’s R2 of 0.156 and a C-Statistic of 0.785 on the validation sample. Please refer to Appendix A (see supplemental attachment pages 27-32) for details on the risk factor coefficients, including standard deviations and p-values.
Calibration and DiscriminationWe calibrated the most recent risk adjustment update by comparing changes in performance for home health agencies overall and by important subgroups (urbanicity/rurality, size, and share of quality episodes with non-white patients) to the prior risk adjustment specification. The results in Figure 3 (see supplemental attachment page 23) below indicate that most home health agencies overall and by subgroup perform equally well based on the updated risk adjustment model compared to the prior model, ranging between 86 percent among urban home health agencies and 88.6 percent among home health agencies with the highest percentage of non-white patients.
Interpretation of Risk Factor FindingsA patient’s improvement in ambulation/locomotion is dependent on a variety of factors, including social risk factors, clinical and behavioral risk factors, and access to care. Risk adjustment is used to promote incentives for home health agencies to provide the same care to patients regardless of patient characteristics at SOC/ROC. All risk factors that are expected to impact the ability to improve in ambulation/locomotion between SOC/ROC and EOC and are based on OASIS items that will remain or will be added following the transition to OASIS-E are considered for model inclusion. The statistical properties of the OASIS items were examined to specify risk factors (e.g., item responses were grouped when there was low prevalence of certain responses). Team clinicians then reviewed all risk factors for clinical relevance and redefined or updated risk factors, as necessary. These risk factors were divided into 31 content focus groups (e.g., functional status, Hierarchical Condition Categories, etc.). The final risk adjustment model was selected by considering model fit statistics, statistical significance of risk factor coefficients, and clinical plausibility of magnitude and direction of coefficients. Please refer to other sections for details on the risk factor selection methodology.
Final Approach to Address Risk FactorsRisk adjustment approachOnRisk adjustment approachOffConceptual model for risk adjustmentOffConceptual model for risk adjustmentOn
- Contributions Towards Advancing Health Equity
Across home health agencies, we compared Improvement in Ambulation/Locomotion (#0167) CY 2022 performance by subgroups for urbanicity/rurality, size, and share of quality episodes with non-white patients (see Figure 4 on supplemental attachment page 24).
We define urbanicity as home health agencies located within a Core-Based Statistical Area (CBSA) as defined by the Office of Management and Budget (OMB). Urban home health agencies perform on average slightly worse at 0.797 than rural home health agencies at 0.810.
We define large home health agencies as home health agencies with quality episode counts in the top quartile for CY 2022 and small home health agencies as being in the bottom quartile. Large home health agencies perform much better at 0.862 than small home health agencies at 0.713.
For “Highest Quartile Non-White” home health agencies, we use the M0140: Race/Ethnicity OASIS item to identify the patient’s race/ethnicity as non-white. Home health agencies in the lowest quartile share of quality episodes with non-white patients perform worse at 0.717 than home health agencies in the highest quartile at 0.817.
The results, particularly for home health agency size and percentage of non-white patients, indicate a performance gap across home health agencies by subgroup. CMS is monitoring the persistence of these gaps and investigating next steps for addressing through reevaluated measure specifications or other policies (see https://www.cms.gov/medicare/quality/home-health-quality-reporting-program/home-health-qrp-health-equity for additional resources).
- Current StatusYesCurrent Use(s)
- Name of the program and sponsorPublic Reporting & Home Health Star Ratings; CMSPurpose of the programQuality improvementGeographic area and percentage of accountable entities and patients includedYes and see belowApplicable level of analysis and care setting
The Home Health Compare website is a federal government website managed by the Centers for Medicare & Medicaid Services (CMS). It provides information to consumers about the quality of care provided by Medicare-certified home health agencies throughout the nation. The measures reported on Home Health Compare include all Medicare-certified agencies with at least 20 home health quality episodes. In the period ending December 31, 2022, there were 7,747 such agencies (78.73 percent of the 9,840 agencies with at least one quality episode) that met the measure denominator criteria for reporting of Improvement in Ambulation/Locomotion. This included 4,443,810 episodes of care nationally.
CMS’s Home Health Quality Initiative provides all Medicare-certified home health agencies with opportunities to use outcome measures for outcome-based quality improvement. The report allows agencies to benchmark their performance against other agencies across the state and nationally, as well as their own performance from prior time periods.
Actions of Measured Entities to Improve PerformanceAll home health agencies with at least 20 qualifying quality episodes of care receive quarterly measure reports on all their publicly reported measures. In addition, providers can run on-demand, confidential reports showing individual measure results and national averages, through CMS’ iQIES system. There is an email box that home health agencies may submit questions to as well as a website on which the latest measure updates are posted. The OASIS Guidance Manual describes the OASIS-based reports that are available, report use(s), and provides guidance about OASIS and quality improvement. Home health agencies make use of these reports to monitor and improve the quality of care.
Feedback on Measure PerformanceHome health agencies receive quarterly measure reports on all their measures. There is an email box that home health agencies may submit questions to as well as a website on which the latest measure updates are posted. Because of the changes made to the OASIS in the OASIS-E version (effectively January 1, 2023), risk models for publicly reported outcome measures have been updated. CMS makes available information about risk models and covariates on its website.
Consideration of Measure FeedbackNo measure specifications changes requested or made.
Progress on ImprovementThe measure is important to report publicly. Although improvements in performance are small, home health agencies continue to improve overall, and for each subgroup measured over time. Performance gaps still exist, and potential future performance metrics are likely to reflect the expansion of the HHVBP Model. Publicly reported measure results illustrate variation in performance across home health agencies that may inform patient and family choice of a home health agency.
Figure 5 (see supplemental page 26) presents trends in risk-adjusted Improvement in Ambulation/Locomotion (#0167) by subgroup. In addition to overall improvement from CY 2019 to CY 2022, each subgroup improves. We expect improvement to be driven in part by the implementation of the Quality of Patient Care (QoPC) Star Rating beginning in July 2015 and the Home Health Value Based Purchasing (HHVBP) Model in 2016. Results prior to 2019 showed dramatic improvement (not shown), while improvement in 2019-2022 was smaller. Nonetheless, QoPC Star Rating and HHVBP still provide incentives for home health agencies to improve on this measure. We anticipate continued improvement as HHVBP expands nationwide in 2023. Data will not reflect this policy change, as we only report results through CY2022.
Unexpected FindingsWe do not find any unexpected findings during implementation of this measure at this time.
-
- Logic ModelPerformance Gap
Improvement in Ambulation/Locomotion (#0167) is calculated using CMS’s Home Health Quality Reporting Program’s assessment tool, the Outcome and Assessment Information Set (OASIS). All components of the measure are defined using data from the OASIS, including the numerator, denominator, exclusions, and risk factors. The measure is risk adjusted to account for patient characteristics at the start of care or resumption of care (SOC/ROC). The denominator consists of unique quality episodes, i.e. a SOC/ROC assessment paired with an end of care (EOC) assessment.
While all the data used to report results in this form are derived from the OASIS, the periods used to generate results vary. Trends are presented from calendar year 2019 (CY 2019) to CY 2022. We restrict descriptive characteristics, reliability, and validity to CY 2022, the most recent calendar year of data currently available. The results generated for risk adjustment use CY 2021 data, the data used during the most recent maintenance reevaluation and risk adjustment update.
Table 2 (see supplemental attachment page 8) presents performance for Improvement in Ambulation/Locomotion (#0167) from CY 2019 to CY 2022 among home health agencies that exceed the public reporting threshold of at least 20 quality episodes of care. Overall, mean performance has been trending upwards, with a low of 0.760 in CY 2019 and a high of 0.798 in CY 2022. The lower and upper bounds of the interquartile range have also increased with each year. Despite the steady increases year-over-year, there remains a performance gap for Improvement in Ambulation/Locomotion (#0167). Fewer than 70 percent of quality episodes exhibit improvement among the lowest quartile of home health agencies, and between 20 and 25 percent of quality episodes fail to improve for the average home health agency (see Tables 2 and 3 on supplemental attachment page 8).
Aside from CY 2020, which was affected by the COVID public health emergency reporting requirements, roughly 4.4 million quality episodes are used to score the measure. The measure is publicly reported for over 7,000 home health agencies.
Table 3 (see supplemental attachment page 8) presents performance for Improvement in Ambulation/Locomotion (#0167) for CY 2022 by home health agency size among home health agencies that exceed the public reporting threshold. From bottom decile to top decile, the distribution in performance is tight with a minimum at Decile 1 of 0.678 and a maximum at Decile 9 of 0.863, a 0.185 difference in mean score. Smaller home health agencies perform worse on the measure, with Deciles 1 to 4 performing lower than the overall mean score and Deciles 5 to 10 performing higher.
- Do you have a secondary measure developer point of contact?OnMeasure Developer Secondary Point Of Contact
Alrick Edwards
Abt Global
5001 S Miami Blvd #210
Durham, NC 27703
United StatesMeasure Developer Secondary Point Of Contact EmailMeasure Developer Secondary Point Of Contact Phone NumberThe measure developer is NOT the same as measure stewardOnSteward Organization EmailSteward Phone NumberSteward AddressKristina Rabarison
CMS
7500 Security Boulevard
Baltimore, MD 21244
United StatesSteward Organization URLSteward Organization CopyrightN/A
- Detailed Measure SpecificationsYesLogic ModelOnImpact and GapYesFeasibility assessment methodology and resultsYesEmpirical person- or encounter-levelNoEmpirical accountable entity-levelYesAddress health equityYesMeasure’s use or intended useYesRisk-adjustment or stratificationYes, risk-adjusted onlyQuality Measure Developer and Steward Agreement (QMDSA) FormThe measure is owned by a government entity; therefore, the QMSDA Form is not applicable at this time.A.10 Additional and Maintenance Measures FormThe Additional and Maintenance Measures Form is not applicable at this time.508 ComplianceOn
-
Enter a comment below