David A. Ganz, MD, PhD

 To compare home health utilization and clinical outcomes between Medicare beneficiaries in the fee-for-service (FFS) and Medicare Advantage (MA) programs, and to compare regional variation.

 We used the 2010 and 2011 Outcome and Assessment Information Set to identify all home health episodes begun in 2010 and to measure 7 clinical home health outcomes that are defined by CMS for public reporting.

 We modeled the probability of home health use, the duration of home health episodes, and each clinical outcome measure as a function of MA versus FFS enrollment and model-specific risk adjustors. Empirical Bayes predictions from generalized linear mixed models were aggregated by hospital referral region (HRR) to create standardized regional measures of home health utilization and mean episode duration

We identified 30,837,130 FFS and 10,594,658 MA beneficiaries (excluding those dually eligible for Medicaid). After adjusting for demographic and clinical patient characteristics, the odds of receiving home health among FFS enrollees were 1.83 times those of MA (95% CI, 1.82-1.84). Adjusted home health duration was 34% longer for FFS (95% CI, 32%-34%). Outcomes differences were small in magnitude and inconsistent across measures. Regional variations in use and duration were substantial for both FFS and MA enrollees. Within HRRs, correlations between FFS and MA utilization rates and between FFS and MA episode durations were 0.51 and 0.94, respectively.

 MA beneficiaries use less home health than their FFS counterparts, but regional factors affect utilization, independent of insurance status.

Home health utilization is a substantial source of Medicare costs; however, whether enrollment in a Medicare Advantage (MA) plan affects home heathcare use is unknown. We compared nationwide and intra-regional use of home health between beneficiaries enrolled in fee-for-service (FFS) Medicare and MA plans. Our findings were as follows:

		FFS beneficiaries used substantially more care, even after adjusting for patient and demographic characteristics.

		Functional outcomes were similar, although FFS episodes were 33% longer in duration.

		Adjusted utilization varied markedly across regions for both FFS and MA, suggesting that MA plans cannot entirely overcome non—health-related regional forces driving use.

Although more than 30% of all Medicare beneficiaries are enrolled in the Medicare Advantage (MA) program,

 studies focusing on the MA patient population are relatively rare, in large part because there are no claims to use as a data source. The capitated payment model introduces a financial incentive to limit wasteful care, whereas none exists in the fee-for-service (FFS) system. However, few published studies have directly measured and compared how much a particular service is used in the 2 systems.

One venue where care is documented for all Medicare beneficiaries is home health. All patients receiving care through CMS-certified home health agencies are required to have a comprehensive clinical assessment at the start and end of care, using the Outcome and Assessment Information Set (OASIS) instrument. For both FFS and MA beneficiaries, there is therefore a record not only that care occurred, but also of the clinical status at the start of care and of the functional outcome. Thus, home health provides a unique window through which the 2 programs can be compared.

The use of home health within Medicare is a topic ripe for review. Home health accounted for more than 5% of all FFS Medicare spending ($18.4 billion) in 2011,

 and an Institute of Medicine (IOM) report identified postacute care as the leading source of unexplained regional variation in healthcare spending via FFS.

 In a report to Congress, the Medicare Payment Advisory Commission (MedPAC) suggested that much of the variation in home health use reflected “fraud, waste, and abuse,” and noted that “the broad program standards and fragmented nature of the FFS program do not encourage effective targeting of the benefit.”

 Whether MA plans, with their incentives to avoid unnecessary care, are more effective at managing home health use is a natural next question. As a first step toward answering it, we compared home health utilization, regional variation, and clinical outcomes among beneficiaries enrolled in the MA and FFS programs.

We identified all home health episodes among Medicare beneficiaries that started in 2010 and ended by December 31, 2011. We first performed a beneficiary-level analysis, comparing the probability (after adjusting for patient characteristics) that an FFS versus an MA beneficiary would have at least 1 home health episode. For those who received home health, we also compared the total number of days enrolled and 7 clinical outcomes that are publicly reported by CMS on the “Home Health Compare” website: improvement in the 3 measures of activities of daily living; acute care hospitalizations; and improvements in pain, dyspnea, and management of oral medications.

 Each of these measures is risk-adjusted (per CMS specifications) using the results of the OASIS assessment at the start of care.

 With the exception of the hospital utilization measure, outcomes are defined by comparing end-of-care and start-of-care OASIS assessments, and are defined only for patients who remain living in the community at the end of the episode.

We first compared each study endpoint as adjusted averages across the FFS and MA populations. Then, to better understand patterns of utilization differences between FFS and MA, we compared FFS and MA beneficiaries within each of the 306 hospital referral regions (HRRs). HRRs were chosen as the geographic unit of interest because the system partitions the country into mutually exclusive regional healthcare markets, which were felt to be the best available approximation for regions serviced by home health agencies.

Using the 100% OASIS file, we grouped start-of-care and end-of-care assessments into episodes according to CMS specifications.

 OASIS episodes represent a complete cycle of care that ends only when a patient is hospitalized, dies, or is discharged to the community. An OASIS episode may encompass 1 or more 60-day billing episodes, as defined by the home health prospective payment system.

 To each OASIS episode (observation), we merged in data from the  100% denominator file and from a file containing Hierarchical Condition Category (HCC) scores (which represent the ratio of an individual’s expected cost to that of an average beneficiary).

The Medicare health insurance claim number or the combination of Social Security number and date of birth were used as unique identifiers.  We also added demographic data from the American Community Survey, using 5-year estimates for 2007 to 2011, by linking via the zip code of each beneficiary.

Because patients who were dually eligible for Medicare and Medicaid may have received home health benefits through Medicaid—which would not have been visible to us&mdash;we excluded dual-eligible beneficiaries from the analysis. All others were included.

We used logistic regression with robust standard errors to model the odds of receiving home health and the odds of each of the 7 clinical home health outcomes. We used a generalized linear model with a log link and Gaussian distribution to model the geometric mean number of days enrolled in home health (for those who received care). For models of utilization, risk adjustors included age, sex, HCC score, and 2 socioeconomic indicators from the 2007 to 2011 American Community Survey: deciles of median household income in the Zip Code Tabulation Area (ZCTA) and the proportion of households in the ZCTA with only 1 resident (among those with 1 or more residents aged ≥65), as surrogates for the beneficiary’s income and the probability that they live alone. For home health duration, we used 2 additional risk adjustors: whether the initial home health episode started after a hospitalization (postacute) and whether it followed a surgical procedure. For each clinical outcome, we used an extensive set of outcome-specific risk adjustors (derived from the baseline OASIS assessment) that is specified by CMS for the Home Health Compare program.

To generate region-specific standardized utilization rates, we used generalized linear mixed models (with logit link for the probability of receiving home health and log link for home health duration) to model home health use as described for the national analysis, but added a random intercept for HRR. We then aggregated recycled empirical Bayes predictions to create a standardized utilization rate (or mean duration) for MA and FFS, for each of the 306 HRRs.

 We presented our results graphically by plotting each region’s FFS utilization rate against the MA utilization rate, and reported the correlation coefficients for MA versus FFS across the 306 HRRs.

In keeping with recent reports by MedPAC and the IOM, we quantified regional variation by reporting the ratio of the 90th percentile versus the 10th percentile of regional utilization.

 We reported these ratios separately for the FFS and MA populations.

After excluding 8,254,838 individuals dually eligible for Medicare and Medicaid, we identified 30,837,130 FFS beneficiaries and 10,594,658 MA beneficiaries. Baseline characteristics and unadjusted home health utilization rates are summarized in 

. Among the general population, MA patients were slightly older, whereas in the population using home health, the opposite was true. HCC scores were higher for MA in both populations.

Of those enrolled in FFS, 6.3% had at least 1 home health episode; of those in MA, 3.9% did. Among FFS beneficiaries who used home health, the average number of enrolled days was 75.8; among MA beneficiaries, it was 56.3 days. Although MA patients using home health had higher HCC scores than their FFS counterparts (suggesting a higher disease burden), they had less baseline disability according to each of 6 different measures at the start-of-care OASIS assessment. Home health more commonly started after a hospitalization and/or a surgical procedure for MA patients.

FFS Versus MA: Adjusted National Estimates

. We estimated that among the general Medicare population, the odds of starting home health during 2010 were 1.82-fold higher for those enrolled in FFS versus MA, after adjusting for patient characteristics and socioeconomic indicators. Among those who did start home health in 2010, we estimated that the FFS patients were enrolled an average of 34% more days than their MA counterparts, after risk adjustment.

The relationship between FFS versus MA, in regard to clinical home health outcomes, was inconsistent, and observed differences were small in magnitude: FFS outcomes were better with regard to pain (odds ratio [OR], 1.08; 95% CI, 1.07-1.09), ambulation (OR, 1.01; 95% CI, 1.00-1.02; 

 = .03), and management of medications (OR, 1.06;  95% CI, 1.04-1.07); and MA outcomes were better with regard to bed transfer (OR, 0.98;  95% CI, 0.97-0.99), dyspnea (OR, 0.99;  95% CI, 0.98-1.00; 

 = .02), and hospital utilization (an adverse outcome) (OR, 1.16; 95% CI, 1.15-1.17). Improved bathing was not significantly related to FFS versus MA enrollment (OR, 1.00; 95% CI, 0.98-1.02). Hospital utilization, which measures whether home health ended in hospitalization rather than discharge to the community, was the outcome with the greatest difference (favoring MA). Of note, this outcome is not adjusted for duration of home health. Because FFS patients—who have longer home health episodes&mdash;spend more time at risk for hospitalization while enrolled, this relationship may be confounded.

Among the 10% of HRRs (ie, 31 of 306 HRRs) with the highest proportions of FFS beneficiaries with 1 or more home health episodes (90th-percentile FFS), we estimated that 9.4% of all FFS beneficiaries used home health, whereas among the 31 HRRs with the lowest proportion (10th-percentile FFS), only 4.2% did. The ratio of the 90th percentile/10th percentile for FFS was, therefore, 2.2. For MA patients, the 90th percentile for use of home health was 6.3%, the 10th percentile was 2.2%, and the ratio was 2.8. The higher 90th/10th percentile ratio for MA means that by this measure, regional variation in home health use was higher for MA patients than for FFS patients.

An analogous evaluation of the duration of home health episodes yielded the following results: in the 90th-percentile HRR (with regard to average episode duration), mean duration was estimated to be 92 days for FFS and 67 days for MA. The 10th percentile was 39 days for FFS and 31 days for MA. The ratios of 90th/10th percentiles were 2.4 for FFS and 2.2 for MA.

 shows the standardized proportions of FFS and MA beneficiaries receiving home health in each of the 306 HRRs. 

 shows the analogous mean (standardized) duration of home health episodes. These figures illustrate the correlation between FFS and MA rates of home health use (Figure 1), and average home health duration (Figure 2) across regions. Correlation coefficients for MA versus FFS across regions were 0.51 for the proportion using home health and 0.94 for mean duration, meaning regions with longer home health episodes among MA patients also tend to have longer episodes among FFS patients and that this relationship exists to a lesser extent with regard to the probability that an MA or FFS beneficiary in a given region will use home health.

We have shown that Medicare beneficiaries enrolled in FFS use substantially more home health than those in MA, both in terms of the proportion of beneficiaries who receive care and the duration of home health among those who do. This is true after adjusting for age, disease burden, and regional differences in household size and income. We found that despite the average home health episode duration being 34% longer for FFS, differences in clinical outcomes were small in magnitude and inconsistent. Of the 7 outcomes that we measured, FFS fared slightly better on 3, MA fared slightly better on 3 others, and there was no difference on 1.

Assertions of “fraud, waste, and abuse” in the home health industry are based on data from the FFS program and have been made in the context of a history of rapid fluctuations in home health use that have closely paralleled changes in home health agency reimbursement policy.

 In 1997, after Medicare spending for home health quadrupled over a 5-year period, cost-based reimbursement was abruptly replaced with a forerunner of today’s home health prospective payment system, whereupon spending and utilization rates were cut in half within a year. Subsequently, utilization and expenditures have again crept upward, and since 2002, total FFS home health expenditures have nearly doubled.

The MA versus FFS comparison of home health utilization across regions is of interest because MA plans, which are paid a capitated rate to cover all needed services, have a direct incentive to minimize financial costs, whereas there is no real equivalent incentive for FFS. It is unclear how financial incentives might affect an MA plan’s “propensity to provide home health.” On the one hand, MA plans might be expected to minimize the direct costs of paying home health agencies and restrict the use of home health. On the other, they might substitute home health for more costly inpatient services and be more inclined to promote its use. Whichever the predominant effect, whether MA enrollees actually use less home health than their FFS counterparts also depends on the proportion of FFS home health that is truly wasteful and the practical ability of MA plans to curb such waste. Small studies from the pre-prospective payment system era suggested that Medicare health maintenance organizations (the forerunner of today’s MA plans) used substantially less home health than their FFS counterparts at the time, and our study confirms that this is still true (albeit to a lesser degree).

Regional variation is of interest because it provides insight into the practical limitations to optimizing the use of a service: if MA plans were perfectly able to optimize health outcomes or their own financial interests, then unexplained regional variation (ie, beyond what can be explained by differences between patients) would not be expected. The strong correlation between the use of home health by MA and FFS patients within regions therefore implies that MA plans’ ability to manage utilization is limited; MA plans limit use, but cannot entirely overcome the same regional forces (eg, healthcare market, local practice styles) that drive variation across regions in the FFS population.

We found a greater correlation between the duration of home health episodes for MA versus FFS patients within a region compared with the regional correlation in the proportion of beneficiaries enrolled in MA versus FFS who receive any home health at all. One explanation might be that home health agencies have more leverage in determining duration of care than they do in determining whether beneficiaries are referred to them. If most home health agencies treat a mix of MA and FFS beneficiaries, and if they provide care in a “payer-blind” way, then this finding might be expected.

There has been little previous study of regional variation in the use of a service by MA beneficiaries. One study did compare the use of cardiovascular procedures across 32 of 306 HRRs and found regional FFS and MA rates to be correlated, albeit less so than we find here.

 An IOM report has also noted regional correlation (0.66) in total per-capita spending.

 Regional variation among the FFS population has been extensively studied, but we found more variation in FFS home health use than has been recently described for most other services or for postacute care in general.

 We noted that postacute care encompasses more than home health and that not all home health is post acute. However, a 1996 study of FFS home health use during an era prior to the prospective payment reform found a degree of regional variation that appears to be on par with what we find here (methodological differences preclude an exact comparison).

We cannot directly determine whether greater home health use by FFS beneficiaries means that they receive unnecessary care or whether MA beneficiaries are not receiving necessary care; it is likely that both are sometimes true (as would sometimes be the converse). Measurement limitations notwithstanding, our finding that MA and FFS patients have similar clinical outcomes suggests that the shorter MA episodes may not be detrimental to health. We do not address patient preferences, however; nor do we explore the clinical effect of differing rates of home health use (compared with differing episode durations).

Our study has several other limitations. First, because MA plans have a financial incentive to select for healthier patients in ways that are not captured by CMS payment models,

 it is possible that unobservable differences between the populations remain after risk adjustment. Although this could account for some of the difference that we see in a beneficiary’s probability of receiving home health, we would expect less influence on our estimates of home health duration or outcomes, which are conditional on receiving home health and which use a larger set of observables (from OASIS) for risk adjustment. Second, functional outcome measures rely on clinical assessments at the start and end of each home health episode. Because each agency’s performance on these measures is publicly reported, there may be an incentive for agencies to underestimate functional status at the start of care or overestimate it at the end of care, biasing the results toward better outcomes, which would have the effect of minimizing any true outcomes differences between the FFS and MA populations. Moreover, since the end-of-care assessment does not take place when patients die or are transferred, outcomes are only measured for patients who are discharged back to the community at the end of home health. Finally, we do not have a direct window onto the actual number of home health visits that a patient receives, and rely instead on the duration of a home health episode as a proxy. We cannot be certain that FFS and MA beneficiaries receive the same number of visits per unit of time enrolled in home health.

MedPAC has reported to Congress that the FFS system is not structured to limit wasteful use of the home health benefit.

 Our study uses MA as a counterexample of care under an alternative reimbursement structure. The Government Accountability Office cites the lack of standards for what constitutes appropriate and necessary care as one of the challenges to curbing excess. Although we do not offer such a standard, we suggest that the type of within-region, between-payer comparison performed here can help to better define a reasonable range of use and to identify high-use and low-use regional outliers (for both FFS and MA). This methodology may be more widely applicable using MA encounter data, which are still being collected.

The authors acknowledge Patty Smith for editorial assistance, Cheryl Damberg and Rodger Madison for assistance with data access, and Emmett Keeler and Jeffrey Wasserman for comments on a previous version of this paper. 

Department of Emergency Medicine (DAW) and David Geffen School of Medicine (NSW, DAG), University of California, Los Angeles, Los Angeles, CA; RAND Corporation (DAW, LM, CMS, MH, NSW, DAG), Santa Monica, CA; University of Michigan Medical School (LM), Ann Arbor, MI; VA Ann Arbor Healthcare System (LM), Ann Arbor, MI; VA Greater Los Angeles Healthcare System (DAG), Los Angeles, CA.

The analyses upon which this article is based were performed under contract HHSM-500-2005-00028I, task order number HHSM-500-T0004, entitled “Evaluation and Development of Outcome Measures for Quality Assessment in MAOs and SNPs,” funded by CMS and HHS. The content of this article does not necessarily reflect the views or policies of HHS, nor does the mention of trade names, commercial products, or organizations imply endorsement by the US government. The author assumes full responsibility for the accuracy and completeness of the ideas presented.

Dr Waxman was supported by the Department of Veterans Affairs (VA) Office of Academic Affiliations through the VA Health Services Research and Development Advanced Fellowship Program; he had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Dr Ganz was part of the project team on contracts awarded to RAND from CMS and the Health Services Advisory Group, and was part of the project team on a contract awarded to UCLA from SCAN Health Plan. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article. 

Concept and design (DAG, MH, LM, DAW); acquisition of data (DAW); analysis and interpretation of data (DAG, MH, LM, CMS, DAW); drafting of the manuscript (LM, CMS, DAW); critical revision of the manuscript for important intellectual content (DAG, LM, DAW); statistical analysis (CMS, DAW); obtaining funding (DAG, DAW); administrative, technical, or logistic support (MH); and supervision (DAW). 

Daniel A. Waxman, MD, PhD, RAND Corporation, 1776 Main St, Santa Monica, CA 90407. E-mail: dwaxman@rand.org.

1. Medicare Advantage. Kaiser Family Foundation website. http://kff.org/medicare/fact-sheet/medicare-advantage/. Published May 11, 2016. Accessed February 13, 2016.

Articles

Medicare Advantage beneficiaries use less home healthcare than do their fee-for-service counterparts, but there is marked regional variation in use by both groups.

Does Medicare Advantage Enrollment Affect Home Healthcare Use?

Although Medicare has implemented incentives for health plans to reduce fall risk, the best way to identify older people at high risk of falling and to use screening results to target fall prevention services remains unknown. We evaluated 4 different strategies using a combination of administrative data and patient-reported information that health plans could easily obtain.

We used data from 1776 patients 75 years or older in 4 community-based primary care practices who screened positive for a fear of falling and/or a history of falls. For these patients, we predicted fall-related injuries in the 24 months after the date of screening using claims/encounter data. After controlling for age and gender, we predicted the number of fall-related injuries by adding Elixhauser comorbidity count, any claim for a fall-related injury during the 12 months prior to screening, and falls screening question responses in a sequential fashion using negative binomial regression models. 

Basic patient characteristics, including age and Elixhauser comorbidity count, were strong predictors of fall-related injury. Among falls screening questions, a positive response to, “Have you fallen 2 or more times in the past year?” was the most predictive of a fall-related injury (incidence rate ratio [IRR], 1.56; 95% CI, 1.25-1.94). Prior claim for a fall-related injury also independently predicted this type of injury (IRR, 1.41; 95% CI, 1.05-1.89). The best model for predicting fall-related injuries combined all of these approaches.

The combination of administrative data and a simple screening item can be used by health plans to target patients at high risk for future fall-related injuries.  

Among a high-risk population of community-dwelling older adults, the best model for predicting future fall-related injuries combined basic administrative data (age, gender, and comorbidity count), a prior claim for a fall-related injury, and patient response to the single item screener, “Have you fallen 2 or more times in the past year?”

			The most important influences on prediction were comorbidity count and response to the screening question. 

			Given current Medicare incentives to reduce fall risk, health plans can use this information to target fall prevention interventions to the highest risk members without requiring a face-to-face office visit.

Reducing fall risk is one of the process measures used to determine the CMS 5-Star Ratings for Medicare Advantage plans, which are published annually, and to determine Medicare Advantage Quality Bonus Payments.

  Moreover, with the passage of the Affordable Care Act, private insurers, Medicare, and Medicaid are likely to be required to provide coverage for new preventive health services graded A or B by the US Preventive Services Task Force (USPSTF),

 including exercise or physical therapy and vitamin D supplementation for fall prevention (USPSTF grade B recommendations).

  However, the best method to screen community-dwelling older adults for fall risk, and the best way to use the screening results to target preventive services for falls, remains unclear.

Several risk factors have been identified for falls among community-dwelling older adults,

 with age and a history of falls being the 2 most commonly used risk factors to define high risk in fall intervention studies.

 Other frequently reported risk factors for falls include a history of mobility problems, poor performance on office-based gait and balance testing such as the Get-Up-And-Go test, visual impairment, use of psychoactive medications, and female gender.

 However, there is no consensus on a best single screening question, set of questions, or clinical tool to reliably identify older adults at increased risk for falls.

Most clinical tools used to assess fall risk are performed in a primary care setting by a clinician,

 requiring a face-to-face office visit, provider training in the use of the screening tool, and provider engagement in the screening process. However, this may not be the preferred approach for a health plan seeking to provide fall prevention services more widely to its at-risk members. Alternative approaches could include screening questions completed over the phone or by mail or the use of health plan administrative data to identify patients with risk factors for falls (eg, age, female gender, certain health conditions) or with a prior claim for a fall or a fall-related injury.

Our objective was to examine the predictive value of different possible screening approaches for fall risk that would be relevant to a health plan seeking to identify its older adult members at highest risk for falls and fall-related injuries. We assessed the value of 3 commonly used falls screening questions in combination with data from Medicare claims in predicting subsequent fall-related injuries in a cohort of high-risk community-dwelling older adults.

We studied 1776 adults 75 years or older who screened positive for falls or fear of falling in 4 community-based primary care practices using 3 questions: 1) Have you fallen 2 or more times in the past 12 months? 2) Have you fallen and hurt yourself since your last visit to the doctor? 3) Are you afraid that you might fall because of balance or walking problems?

Participants were screened during 2006 and 2007 as part of the ACOVEprime study, a controlled trial of a practice-based quality improvement intervention to improve care for falls and urinary incontinence.

This project was approved by the UCLA Institutional Review Board (IRB). The 4 participating sites either approved the project via their own IRB or by deferring to the UCLA IRB. A fifth site received approval to obtain claims only from decedents; thus, data from this site were excluded.

We obtained Medicare claims data for calendar years 2005 to 2009 for all individuals who screened positive for at least 1 of the fall-related screening questions. We were unable to obtain claims data for screen-negative individuals as they were not included in the original ACOVEprime study, and the study sites did not retain registries of screen-negative individuals at the conclusion of ACOVEprime. 

Medicare eligibility was determined based on the Master Beneficiary Summary File; fee-for-service Medicare claims data were obtained from MedPAR, outpatient, carrier, hospice, home health agency, and durable medical equipment files. For participants enrolled in a Medicare Advantage plan at the time of screening, we attempted to obtain both institutional and professional encounter data from participating health plans that had at least 50 members in the study. Of the 2022 individuals who were potentially eligible from the 4 sites, 77 were either not enrolled in Medicare or sites did not have a valid Medicare ID for the patient; we could not obtain managed care data for 169 patients. This left us with a final analytic cohort of 1776 patients.

Our primary outcome measure was the count of fall-related injuries in the 24 months after the date of screening. Since we could not find a claims-based algorithm for fall-related injuries in the literature that was validated against medical records, we adapted an existing approach used for osteoporotic fractures.

International Classification of Diseases, Ninth Revision, Clinical Modification 

) and Healthcare Common Procedure Coding System (HCPCS) codes for hip fractures, other selected nonvertebral fractures, inpatient head trauma, selected joint dislocations, and healthcare claims where an external cause of injury code (“E-code”) for falls was used. A list of injuries with associated 

 and E-codes and the coding hierarchy used has been described in prior work by this group.

 We included claims for fall-related injuries in both the inpatient and outpatient settings in the 24 months after the individual screening date and conducted sensitivity analyses in which we only analyzed the first fall-related injury experienced by each participant. To avoid double counting fall-related injuries due to multiple claims for a single injury, a claim could only count toward defining a unique injury if there were no prior claims for a fall-related injury, or the claim occurred at least 30 days after a previous claim for a fall-related injury. In a robustness check, varying the length of the episode window around 30 days had little effect on the number of unique fall-related injuries identified.

Because the data were obtained from a controlled multi-site study, we began analyses using a model that adjusted for study site and ACOVEprime intervention group status.  We also included a dichotomous variable identifying the few cases (N = 52; 2.9%) where there were missing claims data in the 12-month period prior to the screening date since such cases might have a falsely low rate of comorbidities or claims-detected fall-related injury in the baseline period.  We retained study site, intervention group, and the dichotomous variable adjusting for missing data in all subsequent models as an adjustment for study design.

Descriptive statistics were used to examine age, gender, and comorbidity count derived from Medicare claims using the Elixhauser method

; all of these variables were determined to have a potential relationship with fall-related injuries. We then included variables for age (as a continuous variable) and gender in the model, creating our base model. All additional predictors were added to this base model in a sequential fashion.

We included responses to the 3 falls-screening questions in 2 different ways. First, we included the response (yes/no) to each individual screening question without controlling for responses to the other 2 questions, which allowed us to examine the predictive value of each question individually. Second, we included the screening questions in the multivariate model as a 3-item set, with each response adjusted for response to the other 2 items. Next, we added to our baseline multivariate model a binary variable for the presence/absence of at least 1 fall-related injury during the 12 months prior to the date of screening using our claims-based outcome definition with and without controlling for screening question response. This allowed us to evaluate the utility of a model using only administrative data to predict future fall-related injuries and a model using both administrative and patient-reported information (patient demographics, screening item response, and prior claim for a fall-related injury).

For our primary analyses, we examined the count of fall-related injuries at 24 months after the date of screening using negative binomial regression models. Negative binomial regression was selected as the primary approach to allow all falls, including recurrent falls, in a given participant to be counted in the analyses and to account for the individual follow-up time of each participant based on available claims data.

 Models were compared using Akaike’s information criterion (AIC) and Bayesian information criterion (BIC), likelihood measures in which lower values indicate better fit and a penalty is paid for increasing the number of variables in the model.

 A stronger penalty is paid for increasing the number of variables with BIC.

Lastly, we compared observed and predicted total fall-related injuries by decile of predicted values and conducted Hosmer-Lemeshow χ

 tests to examine goodness-of-fit for all models.

 We also examined scatter plots of residuals versus predicted values with locally weighted polynomial regression (LOESS) smoothing curves for all models.

We also conducted sensitivity analyses examining time to first fall-related injury for each participant during the 24 months after screening using Kaplan-Meier plots and Cox proportional hazards models. As both approaches resulted in similar findings, we present only the negative binomial models here. All statistical analyses were performed using STATA IC version 13 (Stata Corp LP, Texas Station, Texas).

Of the 1776 patients who screened positive for a fear of falling and/or a history of falls who had data available for analysis, the mean age and number of comorbidities were 82.9 years (SD = 5.3) and 2.9 (SD = 2.4), respectively; 72% of participants were women. The majority of the sample reported being afraid of falling due to balance or walking problems (84%). More than one-third reported 2 or more falls in the past year and about one-fourth reported having had a fall resulting in injury since their last visit to the doctor (

 shows incidence rates for fall-related injuries per 1000 person-years before and after screening.  Eleven percent of participants had a claim for a fall-related injury in the 12 months prior to study entry and 19% had any claim for a fall-related injury during the 24-month study follow-up period.

 shows the 4 different models for predicting fall-related injuries. The greatest increase in predictive value for future fall-related injuries occurred with the addition of Elixhauser comorbidity count to the base model with a decrease in AIC of 22.7. Age and Elixhauser comorbidity count remained significant predictors of future fall-related injuries in all models.

Adding the response to the single-item screening question, “Have you fallen 2 or more times in the past year?” increased the predictive value of the model with age, gender, and comorbidity count more than the addition of a prior claim for a fall-related injury, with a decrease in AIC of 18.3 and 7.6, respectively. Adding prior claim for a fall-related injury to the model with patient demographics and response to the single-item screening question was most predictive of future fall-related injuries and further decreased the AIC by 3.1. The comparison of models by BIC was generally similar to the comparison using AIC and is provided in 

). The comparison of models by relative likelihood of better fit is also provided in eAppendix 1. 

In the final model, a 5-year increase in age was associated with a 22% relative increase in the expected number of fall-related injuries in a 2-year period, and each additional comorbid condition was associated with a 10% relative increase in the expected number of fall-related injuries. Observed and predicted fall-related injuries for the final model (Model 4 in Table 3) were similar across deciles of risk, as seen in the 

. Plots of observed and predicted values for Models 1, 2, and 3 in Table 3 and Hosmer-Lemeshow test statistics for all models are provided in 

, and plots of residuals versus predicted values for all models are provided in 

 shows the number of fall-related injuries predicted by the model for an 85-year-old female with different combinations of risk factors. For example, an 85-year-old woman with 3 comorbid conditions and a prior fall-related injury who reports 2 or more falls in the last year is expected to have 0.45 fall-related injuries within the next 2 years. In contrast, another woman of the same age who has none of the 3 risk factors is expected to have 0.15 fall-related injuries within the next 2 years.

“Have you fallen 2 or more times in the past year?” was the only screening question that independently predicted fall-related injuries in multivariate models (for a positive response: incidence rate ratio (IRR), 1.56; 95% CI, 1.25-1.94). Neither, “Have you fallen and hurt yourself since your last visit to the doctor?” nor “Are you afraid that you might fall because of balance or walking problems?” was predictive of a fall-related injury (IRR, 1.15; 95% CI, 0.89-1.48; and IRR, 1.33; 95% CI, 0.98-1.82, respectively) in multivariate models. Prior claim for a fall-related injury also independently predicted future fall-related injuries (IRR, 1.41; 95% CI, 1.05-1.89), after adjusting for patient demographics and screening item response.

Taking the perspective of a health plan, we developed a model to identify patients at high risk for future fall-related injuries using a combination of administrative data and response to a simple single-question screening item that could be administered by mail or phone. The largest gains in prediction were seen with the addition of Elixhauser comorbidity count and response to the single-item screening question, “Have you fallen 2 or more times in the past year?” Adding the response to, “Have you fallen and hurt yourself since your last visit to the doctor?” to the model did not improve the model’s ability to predict fall-related injuries. These findings are consistent with prior work showing that a history of multiple prior falls is a strong predictor of future falls.

 In contrast, fear of falling added little to the regression modelling. This finding is similar to other multivariate analyses,

 where fear of falling was a weakly positive predictor of future falls. 

These findings have practical implications for insurers seeking to identify those members at highest fall risk. Using plan administrative data and a one-item screener, health plans have the ability to use similar models to identify those at highest risk of fall-related injuries. While other multivariate models predicting fall risk exist,

 our approach is novel in that we identify those at highest risk using available administrative data and without requiring a face-to-face office visit.  We used age, gender, and Elixhauser comorbidity count to adjust for patient demographics and global disease burden in our models; however, health plans could use other conceptually similar markers of disease burden, such as the CMS Hierarchical Condition Categories (HCC) risk score, which uses age, gender, prior year diagnoses from Medicare claims, Medicaid dual eligibility, disability status, and institutionalization to risk-adjust Medicare program payments.

  Although a prior claim for a fall-related injury independently predicted a future incident, including this variable only modestly improved model fit and also would require a health plan to undertake operationalizing a claims-based algorithm for determining fall-related injuries. â€©There are several different ways a health plan could use this combination of administrative and questionnaire data to target fall-prevention interventions. Multifactorial falls reduction programs that include physical therapy designed to reduce fall risk

 could be directly offered to those members at highest risk. Alternatively, a fall-risk profile could be provided to the primary care provider to alert him or her that the patient may need further evaluation or a more individualized intervention to prevent future falls. 

How a health plan chooses to implement screening for fall risk also depends in part on the available resources dedicated to screening and implementation of prevention activities. While contacting patients individually to ask about prior history of falls is more time- and resource-consuming than using administrative data alone, the responses to this screening question also more accurately identify those likely to utilize healthcare services for falls or fall-related injuries. Alternatively, a model that uses administrative data entirely—particularly one that includes a measure of global disease burden&mdash;can still reasonably identify those at risk for future fall-related injuries, but would encompass a broader population including those at lower risk. 

This study evaluates different strategies to identify high fall risk among community-dwelling elders using a combination of easily obtainable administrative data and patient screening questions. 

Targeting a High-Risk Group for Fall Prevention: Strategies for Health Plans

 To determine whether nurse practitioner (NP) comanagement can improve the quality of care for 5 chronic conditions in an academic geriatrics practice.

 From September 2006 to September 2007, 18 primary care geriatricians were divided into an intervention group that could refer patients to an NP for comanagement of dementia, depression, falls, heart failure, and/or urinary incontinence, or a control group that indicated which patients would have been referred to the NP for these conditions. The NP used structured visit notes to guide care delivery for the 5 conditions concordant with Assessing Care of Vulnerable Elders-3 (ACOVE-3) quality indicators. We reviewed charts to determine adherence to recommended processes of care.

 A total of 200 patients (108 intervention, 92 control) were eligible for at least 1 process of care recommended by ACOVE-3 for the 5 conditions. Patients' mean (SD) age was 85 years (7 years), 67% were women, and patients were eligible for a mean (SD) of 6.9 (4.4) processes of care. Intervention patients were eligible for more care processes than controls (7.8 vs 5.9 processes per patient; 

 = .002). Quality of care was higher for patients in the intervention group compared with the control group (54% vs 34% of care processes completed; 

 <.001). The adjusted absolute difference between intervention and control groups in care processes completed was 20% (95% confidence interval = 13%, 27%).

 NP comanagement of 5 chronic conditions was associated with higher quality of care, even in a practice of geriatricians.

	A nurse practitioner markedly improved the quality of care for dementia, depression, falls, heart failure, and urinary incontinence in an academic geriatrics practice.

		Adding a nurse practitioner focused on managing these 5 chronic conditions improved adherence to recommended geriatric care by 20% in a setting where all physician providers were geriatricians.

		Improved quality of care came at the expense of 2.6 additional visits to the nurse practitioner per patient during a 13-month period.

	The United States needs a more robust primary care system, yet fewer medical students are pursuing careers in primary care.

 This quandary, coupled with the need for interprofessional collaboration to care effectively for older adults,

 has renewed interest in care models that maintain the primary care physician (PCP) as the key decision maker with patients, but involve others in care implementation, including registered nurses with additional training (eg, nurse practitioners [NPs]), social workers, and physician assistants.

 Data suggest that NPs focused on a particular condition can provide at least comparable quality of care for that condition and sometimes achieve better patient  outcomes.

 However, informationis sparse regarding the effectiveness of NP comanagement of complex  older patients in outpatient settings. An NP/social worker team collaborating with PCPs and a geriatrician-led interdisciplinary team improved the quality of medical care for low-income seniors,

 and a trial of a registered nurse working with    teams of PCPs to improve care for older adults with multimorbidity has shown an early beneficial effect on quality.

 In addition, specialized nurses, NPs, or psychologists have successfully comanaged patients with depression or dementia in primary care.

	Available studies do not, however, address whether NP comanagement can improve quality in a primary care practice staffed by geriatricians. By virtue of their training, interest, and patient population served, geriatricians are likely to be expert in managing geriatric conditions. Yet many of the barriers to providing recommended care that affect all PCPs (eg, lack of time during the office visit, inability to overcome existing practice habits)

 also confront geriatricians. This study’s goal was to evaluate a quality improvement program that  compared usual primary care by academic geriatricians with care comanaged by an NP for 5 chronic conditions: dementia, depression, falls, heart failure, and urinary incontinence.

	The quality improvement project ran from July 1, 2006, to September 30, 2007. Subsequently, University of California, Los Angeles Institutional Review Board approval was obtained to examine de-identified data. 

 contains a description of the project setting.

	The project consisted of run-in (July 1, 2006 to August 30, 2006) and intervention (September 1, 2006 to September 30, 2007) periods. For these periods, an experienced, gerontology-certified NP (JKB), who was hired specifically for the comanagement role, was supported by a grant from the John A. Hartford Foundation. The NP was colocated with the geriatricians at the large outpatient clinic (see Table 1 for details on practice locations).

	During the run-in period, the NP customized condition-specific structured visit notes based on previous versions

 to address the Assessing Care of Vulnerable Elders-3 (ACOVE-3) quality indicators for the 5 chronic conditions

 and American College of Cardiology/American Heart Association heart failure guidelines.

 These note templates were designed to support Medicare documentation requirements and could be used for both initial and follow-up visits. The run-in period also allowed time for the NP to develop her practice style, familiarize herself with clinic operations, and build trust among the referring physicians while seeing patients together with her supervisor (BKK). The practice also developed condition-specific patient education handouts (based on published materials

) and referrals to local community resources, most of which were available to both intervention and control physicians.

	Before the intervention period, physicians were first paired by patient volume; then 1 physician from each pair was selected by coin toss to be able to refer to the NP. Intervention physicians could refer only their primary care patients, not other physicians’ patients whom they might have seen while covering. Control physicians could not refer patients to the NP but were requested to indicate which patients they would have referred.

	During the intervention period, the night before a physician’s scheduled clinic, practice staff clipped an additional sheet to each patient’s chart. Control physicians were asked to note which (if any) of the 5 conditions would have prompted referral of the patient for NP comanagement and the priority level for each “referred” condition (more than 1 condition could have the same priority level). Intervention physicians were asked to note whether they would like the patient to see the NP for any of the 5 conditions and the priority level for each referred condition. The clinic scheduler received completed forms and, in the case of intervention patients, arranged an appointment with the NP. The NP could order tests and treatments without approval from the patient’s geriatrician, but could obtain guidance if needed. The NP delegated certain tasks to clinic nurses, including orthostatic vital signs, urinalyses, bladder scans, and visual acuity measurement. Patient follow-up visits were scheduled by the NP as needed. After each visit, the NP e-mailed or faxed geriatricians a written assessment and plan; she handled more urgent issues via phone or pager.

	The project was introduced at a faculty meeting in September 2006, with follow-up by BKK (the clinical champion) to reinforce the importance of referrals. In February 2007, one pair of high-volume physicians switched roles, with the intervention physician switching to the control group and vice versa, because the physician originally assigned to the intervention group was not referring patients to the NP. This switch was made to increase the number of NP referrals.

	For participating intervention and control physicians who were internists, the practice successfully applied to the American Board of Internal Medicine to receive credit for the Practice Improvement Module of the Maintenance of Certification program on the basis of this project. After the project period, all geriatricians could make referrals to the NP.

	After project completion, 2 abstractors, who had worked as physicians outside the United States, were trained by a nurse researcher with extensive experience in chart abstraction (CPR). Abstractors reviewed both the paper and electronic medical record using structured forms. All intervention and a sample of control group records were reviewed (see 

 for details on control group selection).

	The chart abstraction focused on determining eligibility for, and completion of, care processes specified by relevant ACOVE-3 quality indicators for vulnerable elders, the population served by the practice (see 

for indicator list). The quality indicators specify a time window during which relevant care should be completed; if this window extended beyond the end of the project period for a particular indicator, the patient was considered ineligible for that indicator. The abstractors’ decision about completion of recommended   processes of care was based on the entire chart, regardless of who completed the care process or for which conditions the patient was referred to the NP. Thus, the care reviewed for patients in the intervention group was a compriorbination of both physician and NP care, and included project conditions that physicians had chosen not to refer to the NP. Abstractors were not blind to intervention versus control group status (because structured visit notes appeared only in intervention patients’ records) but were blind to which    conditions patients were referred for.

	Fifteen charts were reabstracted (8 for test-retest reliability and 7 for inter-rater reliability), demonstrating a pooled kappa22 of 0.64 for agreement on eligibility for care processes and (where agreement existed on eligibility) 0.31 for completion of care processes. Data from the paper abstraction forms were entered electronically (with more than 99% accuracy). Abstractors also reviewed encounter data in the electronic health record to determine the number of geriatrician and NP office visits in the intervention and control groups.

	Patient demographics were limited to information in the chart, which included age and sex. Patients were assigned to the intervention or control group based on the group status of their primary geriatrician. For the 2 physicians who switched intervention/control group status during the project, analysis was based on actual referrals (intervention or control) at the time of referral rather than intention to treat. Adherence to each recommended care process was scored dichotomously (process completed/not completed). The percent adherence was computed for the intervention and control groups, both in aggregate and stratified by condition. The primary analysis included all of the patient’s conditions. In a sensitivity analysis, we included only those conditions designated as “top priority” for referral to the NP on the referral sheets. Details on statistical analyses may be found in 

	Charts from 275 patients (139 intervention, 136 control) were reviewed. For 2 patients, no record was found of any geriatric outpatient care; these patients were excluded. Of the remaining 273 patients, 200 (73%, 108 intervention, 92 control) were eligible for at least 1 ACOVE-3 process of care; these patients were eligible for a total of 1384 care processes (mean of 6.9 processes per patient [SD 4.4 processes]). Patients’ mean age was 85 years (SD 7 years); 67% were women. Age and sex did not differ by intervention versus control group assignment (

 = .51, respectively). Intervention patients were eligible for more processes of care than controls (7.8 vs 5.9 processes per patient; 

	Quality of care was higher for the intervention group than for the control group (54% vs 34% of care processes completed;

 <.001). The adjusted difference in quality of care  between intervention and control groups remained significant (

<.001), and the bootstrapped quality of care difference between groups was 20% favoring the intervention group (95% confidence interval = 13%, 27%). Quality of care was higher in the intervention group for all 5 conditions, and the improvement was statistically significant for dementia, falls, and incontinence (

). When we restricted quality-of-care data to those conditions designated by physicians as top prior ity for referral to the NP (found in 148 patients, 74% of  the sample eligible for at least 1 care process), quality of care was 60% in the intervention group and 38% in controls (

 <.001). The intervention group had higher quality of care across all 5 conditions, and the improvement was statistically significant for dementia, falls, and heart failure (Table 2).

	Encounter data were available for 273 patients (99%, 138 intervention and 135 control). During the project period, intervention patients made a mean of 2.5 (SD 3.2) visits to the NP. There was no difference in the number of other visits to the practice by group (7.2 intervention, 6.7 control; 

 =.32). Among those patients (73%, 108 intervention and 92 control) who were eligible for at least 1 ACOVE-3 process of care, intervention patients made a mean of 2.6 (SD 3.3) visits to the NP. In this subsample, other visits to the practice did not differ (6.9 intervention, 6.7 control; 

	When compared with usual care by academic geriatricians, NP comanagement of 5 chronic conditions with geriatricians correlated with an absolute 20% increase in adherence to recommended care. Care for all 5 conditions was better, and this improvement was statistically significant for dementia, falls, and incontinence. Thus, NP comanagement appears to augment the quality of geriatric care even with geriatricians. Comanagement improved quality with a mean of 2.6 NP office visits during a 13-month period.

	These findings extend prior research. Among low-income seniors, one study found better quality of care for depression, urinary incontinence, and falls (among other conditions)  using an NP and social worker who provided home-based care management, together with a geriatrics interdisciplinary team.

 Our project differed in that we provided a less    intense intervention (the NP saw patients in the office and followed up with them by phone rather than making home visits), and our intervention did not target low-income patients. A test of primary care comanagement using specially trained registered nurses showed an early beneficial effect on quality; more definitive results will be reported in the future.

	Strikingly, the quality of care for chronic conditions received by control patients in this project was similar to that received by patients in previous studies who were cared for primarily by nongeriatricians, as measured by ACOVE-3 indicators.

 Differences from previous studies in the methods used to identify conditions and to abstract medical records, as well as the evolution of quality indicators over time, are potential explanations. Alternatively, the results may indicate that physicians face similar barriers to providing high technical quality of care for older adults, regardless of specialty.

	This analysis has limitations. First, intervention group patients were eligible for more quality indicators than controls, suggesting nonequivalence of the 2 groups. Because sicker patients may receive better quality of care

and because being eligible for more quality indicators may indicate a higher level of disease burden, results may be biased in favor  of the intervention group. However, this potential bias is unlikely to account for the entire intervention effect. Second, 1 pair of physicians switched intervention/control group status during the project. We believe that the physician unwilling to refer patients to the NP felt total responsibility for patients and was unwilling to delegate care for this reason. Because the switch occurred to increase referrals to the NP, data were analyzed accounting for the switch rather than as intention to treat. Third, about 25% of patients were not eligible for ACOVE-3 quality indicators. This finding reflects the restricted scope of the ACOVE-3 quality indicators and potentially sparse clinical documentation that limited verification of patients’ eligibility for quality indicators.

	Fourth, enhanced documentation by the NP in the intervention group may have facilitated abstractors noting completion of certain processes of care; however, prior work (not specifically including NPs) suggests that documentation is not the sole basis of poor quality of care.

 Fifth, reliability of chart reabstraction was only fair for completion/ noncompletion of recommended care processes. However, the resulting random misclassification would make estimates of between-group differences conservative. Finally, this project may be a best-case scenario because several of the study investigators, including the NP, also were clinicians in the improvement project and were motivated for the project to be properly implemented.

	This study has significant strengths, however. First, external validity was improved by the quality improvement design, because all patients were eligible for participation as long as they had 1 of the 5 chronic conditions. Second, the intervention was relatively simple and can be replicated with an experienced, motivated NP using materials in the public domain.

	In conclusion, a quality improvement project using an NP to manage 5 chronic conditions was associated with markedly higher technical quality of care for these conditions,  suggesting that deploying NPs as chronic care clinicians may successfully complement physician practice and enhance care.

	We acknowledge all the physicians and staff in the UCLA Division of Geriatrics who participated in this project. Weijuan Han, MSPH, provided programming support. Shakiba Mobaraki, MD, and Farzam Hariri, MD, MPH, did the chart abstraction. We thank the 2 anonymous peer reviewers for their comments.

 From Division of Geriatric Medicine and Gerontology (DAG, JKB, HEM, DBR) and Division of General Internal Medicine (NSW), University of California, Los Angeles; Health Services Research & Development Center of Excellence (DAG), Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, CA; and RAND (CPR), Santa Monica, CA.

Nurse practitioner comanagment improved quality of care for 5 chronic conditions in an academic geriatrics practice.

David A. Ganz, MD, PhD

Articles

Does Medicare Advantage Enrollment Affect Home Healthcare Use?

Targeting a High-Risk Group for Fall Prevention: Strategies for Health Plans

Nurse Practitioner Comanagement for Patients in an Academic Geriatric Practice