Abstract
Background:
Frailty is a geriatric syndrome with negative health impacts not captured by comorbidity and disability alone. The prevalence of frailty in Parkinson disease has been described, but data on frailty-associated outcomes are limited.
Objective:
To describe the level of frailty and investigate the association between frailty and outcomes in a Medicare sample of persons diagnosed with Parkinson disease.
Methods:
We used the claims-based frailty index to assess frailty in a cohort of Medicare beneficiaries with Parkinson disease in 2013. Frailty was categorized as non-frail/prefrail, mildly frail, moderately frail, and severely frail. Adjusted logistic regression models examined the relationship between frailty and mortality, hospitalization, emergency department visits, and fall-related injuries through 2014.
Results:
Of 62,786 beneficiaries with Parkinson disease in 2013, 55.3% were frail. Frail individuals were more likely to be female, older, Black, metropolitan dwelling, without neurologist care, nursing facility residents, or multimorbid. The average daily levodopa equivalent dose initially increased, then decreased from the prefrail to the severely frail groups. Compared to non-frail/prefrail persons, severely frail persons had higher adjusted odds of one-year mortality (AOR=2.74, 95% CI: 1.98, 3.78), hospitalization (AOR=2.34, 95% CI: 1.74, 3.14), emergency department visits (AOR=2.97, 95% CI: 2.14, 4.13), and fall-related injury (AOR=1.43, 95% CI: 0.90, 2.26).
Conclusion:
Frailty is common and differentially distributed among older adults with PD. Frailty in PD is associated with adverse health outcomes and death. Observational study analyses may benefit from adjustment for frailty; claims-based frailty surveillance may identify vulnerable PD patients in health system, registry, or administrative data.
Keywords: Parkinson Disease, Medicare, Frailty, Outcomes, Research Methods
INTRODUCTION
Frailty is a state of dwindling physiological reserve and subsequent physiologic vulnerability typically seen with older age.1,2 Frailty impacts, on average, 17% of community-dwelling older adults (≥50 years of age); however, the estimated prevalence differs by setting, population characteristics, and approach to frailty measurement.3 Treatment patterns may vary between frail and non-frail individuals.4 Falls, impairment, disability, hospitalization, and mortality are all associated with frailty.2,5-7 These findings persist even after accounting for comorbidities and disability.2 Given its strong association with adverse health outcomes, failing to adjust for frailty in observational studies of older adults may lead to biased study estimates due to residual confounding.4 Frailty may also function as an effect measure modifier.4,8
Previous studies of persons with Parkinson disease (PD) find a prevalence of frailty between 3.4% and 84%9-23 and that PD pathology is associated with progression to frailty.18 These are noteworthy findings; however, the current literature on frailty in PD is primarily based on data that are drawn from small10-13,15-24 or academic center 14,15,17,19,21,23 samples. Prior work has demonstrated that academic center limited sampling may systematically exclude individuals with PD who are older, female, from minority race/ethnic groups, rural dwelling, or from lower socioeconomic strata.25 Medical claims provide an opportunity to examine frailty and its associated outcomes in a more diverse PD patient sample.26,27
Several instruments have been developed to capture frailty using modalities such as patient questionnaires, clinical examination, or medical claims.7,8,28 Physical frailty phenotype centered instruments characterize frailty based on physical symptoms that may overlap with presentations of PD itself, including unintentional weight loss, weakness, low energy, low physical activity, and slowness.2,9,24 In contrast, deficit- accumulation based instruments assess frailty by summing diagnoses of individual chronic conditions and other ailments.29 We used a validated deficit- accumulation based frailty instrument to 1) describe the distribution of frailty and 2) determine if frailty is associated with key health outcomes among older adult Medicare beneficiaries with PD.
METHODS
Standard Protocol Approvals and Patient Consents
The University of Pennsylvania Human Research Protections Office and the Centers for Medicare and Medicaid Services (CMS) approved this study with a waiver of individual level consent.
Data Source
Data were obtained from the 2011-2014 CMS Research Identifiable Files.30 We extracted beneficiary demographics, Medicare enrollment information, and data on specific comorbid conditions from the Chronic Conditions Data Warehouse (CCW) and Master Beneficiary Summary File. All other diagnoses and procedures were extracted from the Carrier file. Prescription claims were identified in the Part D Drug Event file. As detailed below, data beginning in the year 2011 was used to establish Medicare eligibility, PD diagnosis, comorbidities, and treatment; frailty status was measured over the year 2013, and study outcomes were measured from January 1, 2014, through December 31, 2014.
Study Sample
To be included in the cohort study and ensure adequate reference periods, a beneficiary must have (1) qualified for Medicare benefits by being age 65 on or before January 1, 2011, (2) had complete Medicare coverage (Part A and B coverage, no health maintenance organization [HMO] coverage) through the end of 2014 or month of death in 2014, and (3) had Part D prescription coverage with no alternative insurance or prescription programs (e.g., dual eligibility or retiree drug subsidy) over the same period.
We further restricted our sample to individuals under active management for PD. We required beneficiaries to have at least one physician or advanced practice provider (APP) claim for PD (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 332.0) in 2013. We excluded those individuals who also had physician or APP claims for atypical parkinsonian syndromes, including drug-induced parkinsonism (ICD-9-CM 332.1), atypical parkinsonism (ICD-9-CM 333.0, 094.82), dementia with Lewy bodies (ICD-9-CM 331.82), amyotrophic lateral sclerosis (ICD-9-CM 335.20, 335.21) or schizophrenia (per the CCW) at any time between 2011 and 2014.31 We required at least one prescription fill for a U.S. Food and Drug Administration (FDA)-approved antiparkinson medication within six months of a qualifying 2013 PD diagnosis.32
Independent Variables
Our primary independent variable was frailty, which we measured using the claims-based frailty index (CFI).33 The CFI was developed and validated for use in Medicare claims data and differentially weights categories of ICD diagnosis; Healthcare Common Procedure Coding System (HCPCS); and Current Procedural Terminology, 4th Edition (CPT-4) codes.5,33,34 Additional details of the instrument's operationalization with ICD, HCPCS, and CPT-4 codes, as well as weighting procedures, are available from Kim et al. (2018).33 CFI scores were estimated using Carrier file diagnosis and procedure data from January 1, 2013, through December 31, 2013, and each person in our sample was assigned to one of five CFI frailty categories: "non-frail" (CFI score <0.10), "prefrail" (0.10-0.19), "mildly frail" (0.20-0.29), "moderately frail" (0.30-0.39), and "severely frail" (≥0.40).35,36
Outcomes
Our primary outcomes have been examined in prior frailty studies—mortality, hospitalization, emergency department (ED) visits, and fall-related injuries.7 The annual number of acute hospitalizations and ED visits and date of death were extracted from the 2014 Master Beneficiary Summary File.37 We identified 2014 fall-related injuries using claims containing e-codes (E880, E881, E882, E884, E885, and E888)38 in the Carrier file. However, because we only had detailed claims information for beneficiaries in the years that they also received care for PD, we created an injury subset sample limited to individuals who also had a claim for PD in 2014. All outcomes were dichotomized.
Covariates
All covariates were based on 2013 data. Sociodemographic covariates included sex, race (White, Black, Other/Missing [Asian, Hispanic, North American Native, Other, Unknown]), age group, and geographical residence. Geographical residence (metropolitan, non-metropolitan) was determined by classifying beneficiary state and county codes according to USDA Rural-Urban Continuum Codes (RUCCs).39 The 73 beneficiaries with missing RUCC information were assigned to the most common classification—metropolitan. Nursing facility residence was determined by the presence of a CPT-4 code between 993.01-993.13. or 993.1840,41 Chronic condition burden was determined using the Combined Comorbidity Score (CCS).42
Individuals with their first physician or APP diagnosis of PD in 2013 were considered incident PD cases. Beneficiaries with at least one claim submitted by a physician with a neurology specialty code were deemed to have received neurologist care. Deep Brain Stimulation (DBS) surgery was determined by any CPT-4 codes for existing or new DBS procedures (618.55, 618.62, 618.63, 618.64, 618.65, 618.67, 618.68, 618.80, 618.85, 618.86, 618.88, 959.61, 959.62, 959.70, 969.72, 959.3, 959.78, 959.79) in 2011, 2012, or 2013. Dopaminergic medication prescription fills data were converted to daily Levodopa Equivalent Doses (LED) based on a previously published conversion formula.43 The mean daily LED for each patient was calculated based on days of pill coverage for 2013.
Statistical Analysis
All analysis was conducted in SAS v9.4 (Cary, NC). Descriptive analyses were conducted on sociodemographic and clinical characteristics. Differences in beneficiary characteristics by frailty status were assessed with Chi-squared tests and Kruskal-Wallis tests. The association between 2013 frailty and 2014 outcomes was assessed with both crude and adjusted logistic regression models. Models were first adjusted for sociodemographic characteristics (age category, sex, race, and residence). Second, models were adjusted for all covariates described above. Due to sparse numbers, the non-frail and prefrail groups were collapsed for regression analyses.
RESULTS
Study Sample Characteristics
A total of 6,887,060 beneficiaries met Medicare eligibility criteria. Of those, 88,216 (1.3%) met the study criteria for actively treated PD in 2013. After excluding individuals also diagnosed with atypical parkinsonism and other alternative neurodegenerative disease diagnoses, the final sample included 62,786 Medicare beneficiaries with PD (55,568 met inclusion criteria for the injury subset analysis). Table 1 displays beneficiary characteristics for the full and injury subset analysis. Overall, 20.7% (n=13,005) were incident PD cases. In general, individuals in the final study sample were predominantly male, 75-84 years of age, White, metropolitan-residing, and under neurologist care. DBS was uncommon (3.2%, n=1,986), and 13.2% (n=8,307) of the sample resided in a nursing facility at some point in 2013. The median CCS was 1.0 (interquartile range [IQR]=3.00).
Table 1.
Descriptive sociodemographic and clinical characteristics of eligible Medicare beneficiaries with active management of Parkinson disease (PD) in 2013 (n=62,786)
| Characteristic | Full Sample (n=62,786) |
Injuries Subset Analysis* (n=55,568) |
|---|---|---|
| Sex, n (Col %) | ||
| Male | 33,178 (52.8) | 29,862 (53.7) |
| Female | 29,608 (47.2) | 25,706 (46.3) |
| Age Category, n (Col %) | ||
| 65-74 years | 18,022 (28.7) | 16,573 (29.8) |
| 75-84 years | 30,190 (48.1) | 26,931 (48.5) |
| ≥85 Years | 14,574 (23.2) | 12,064 (21.7) |
| Race, n (%) | ||
| White | 60,108 (95.7) | 53,221 (95.8) |
| Black | 1,046 (1.7) | 894 (1.6) |
| Other/Missing | 1,632 (2.6) | 1,453 (2.6) |
| Residence, n (Col %) | ||
| Metropolitan | 48,522 (77.3) | 43,428 (78.2) |
| Non-Metropolitan | 14,264 (22.7) | 12,140 (21.9) |
| PD Status | ||
| Incident | 13,005 (20.7) | 9,678 (17.4) |
| Prevalent | 49,781 (79.3) | 45,890 (82.6) |
| Neurologist Care, n (Col %) | ||
| Yes | 51,529 (82.1) | 47,007 (84.6) |
| No | 11,257 (17.9) | 8,561 (15.4) |
| Nursing Facility Residence, n (Col %) | ||
| Yes | 8,307 (13.2) | 6,622 (11.9) |
| No | 54,479 (86.8) | 48,946 (88.1) |
| DBS, n (Col %) | ||
| Yes | 1,986 (3.2) | 1,900 (3.4) |
| No | 60,800 (96.8) | 53,668 (96.6) |
| Combined Comorbidity Score, Median (IQR) | 1.0 (3.00) | 1.0 (3.00) |
| Mean Daily LED (mg), Median (IQR) | 410 (354) | 431 (381) |
| Frailty Status, n (Col %) | ||
| Non-Frail | 67 (0.1) | 64 (0.1) |
| Prefrail | 28,011 (44.6) | 26,024 (46.8) |
| Mildly Frail | 29,009 (46.2) | 25,185 (45.3) |
| Moderately Frail | 5,495 (8.8) | 4,156 (7.5) |
| Severely Frail | 204 (0.3) | 139 (0.3) |
DBS=Deep Brain Stimulation, LED=Levodopa Equivalent Dose
Subset sample analysis was restricted to individuals with any PD claim in 2014
Frailty
CFI analysis revealed that only 0.1% (n=67) of beneficiaries in our PD sample were non-frail, 44.6% (n=28,011) were prefrail, 46.2% were mildly frail (n=29,009), 8.8% were moderately frail (n=5,495), and 0.3% (n=204) were severely frail (Table 1 & Table 2). Frailty was more common in females, older age groups, Blacks, metropolitan dwellers, incident cases, those without neurologist care, those receiving nursing facility care, and those with more comorbidities (Table 2). The median average annual daily LED was 410 mg/day (IQR=354); 97.0% of beneficiaries (n=60,866) had an average annual LED greater than 0, suggesting treatment with a dopaminergic PD medication. Median annual average LED increased initially, in the non-frail (400 mg/day [IQR=450]) and prefrail groups (439 mg/day [IQR=390]), but declined thereafter (mildly frail=400 [IQR=337], moderately frail=334 [IQR=354], severely frail=304 [IQR=296]). A total of 32 individuals, mostly in the prefrail or mildly frail group had extreme mean daily LED values greater than or equal to 3,000 mg/day (annual mean range=3,000-7,092). Excluding these individuals made no changes in the median values.
Table 2.
Distribution of sociodemographic characteristics, clinical characteristics, and 2014 outcomes by 2013 frailty status among eligible Medicare beneficiaries with active management of Parkinson disease (PD) (n=62,786)
| Characteristic | Non-Frail* (n=67) |
Prefrail* (n=28,011) |
Mildly Frail (n=29,009) |
Moderately Frail† (n=5,495) |
Severely Frail† (n=204) |
p-value (Chi-Squared test) |
|---|---|---|---|---|---|---|
| Sex, n (Col %) | <0.0001 | |||||
| Male | 16,444 (58.6) | 14,328 (49.4) | 2,333 (42.5) | 73 (35.8) | ||
| Female | 11,634 (41.4) | 14,681 (50.6) | 3,162 (57.5) | 131 (64.2) | ||
| Age Category, n (Col %) | <0.0001 | |||||
| 65-74 years | 10,798 (38.5) | 6,467 (22.3) | 729 (13.3) | 28 (13.7) | ||
| 75-84 years | 13,321 (47.4) | 14,288 (49.3) | 2,489 (45.3) | 92 (45.1) | ||
| ≥85 Years | 3,959 (14.1) | 8,254 (28.5) | 2,277 (41.4) | 84 (41.2) | ||
| Race, n (Col %) | 0.0095 | |||||
| White | 26,827 (95.5) | 27,817 (95.9) | 5,464 (95.9) | |||
| Black | 445 (1.6) | 496 (1.7) | 105 (1.8) | |||
| Other/Missing | 806 (2.9) | 696 (2.4) | 130 (2.3) | |||
| Residence, n (Col %) | <0.0001 | |||||
| Metropolitan | 51 (76.1) | 20,997 (75.0) | 22,702 (78.3) | 4,600 (83.7) | 172 (84.3) | |
| Non-Metropolitan | 16 (23.9) | 7,014 (25.0) | 6,307 (21.7) | 895 (16.3) | 32 (15.7) | |
| PD Status, n (Col %) | <0.0001 | |||||
| Incident | 11 (16.4) | 4,664 (16.7) | 6,565 (22.6) | 1,682 (30.6) | 83 (40.7) | |
| Prevalent | 56 (83.6) | 23,347 (83.4) | 22,444 (77.4) | 3,813 (69.4) | 121 (59.3) | |
| Neurologist Care, n (Col %) | <0.0001 | |||||
| Yes | 24,041 (85.6) | 23,167 (79.9) | 4,161 (75.7) | 160 (78.4) | ||
| No | 4,037 (14.4) | 5,842 (20.1) | 1,334 (24.3) | 44 (21.6) | ||
| Nursing Facility Residence, n (Col %) | <0.0001 | |||||
| Yes | 448 (1.6) | 4,349 (15.0) | 3,325 (60.5) | 185 (90.7) | ||
| No | 27,630 (98.4) | 24,660 (85.0) | 2,170 (39.5) | 19 (9.3) | ||
| DBS, n (Col %) | 0.0077 | |||||
| Yes | 877 (3.1) | 968 (3.3) | 141 (2.5) | |||
| No | 27,201 (96.9) | 28,041 (96.7) | 5,558 (97.5) | |||
| Combined Comorbidity Score, Median (IQR) | 0 (0.00) | 0 (1.00) | 2.0 (3.00) | 5.0 (4.00) | 8.0 (3.00) | <0.0001 § |
| Mean Daily LED, Median (IQR) | 400 (450) | 439 (390) | 400 (337) | 334 (354) | 304 (296) | <0.0001 § |
| Death, n (Col %) | 1,276 (4.5) | 3,593 (12.4) | 1,518 (27.6) | 84 (41.2) | <0.0001 | |
| Hospitalization, n (Col %) | 5,081 (18.1) | 9,395 (32.4) | 2,475 (45.0) | 117 (57.4) | <0.0001 | |
| Emergency Department Visit, n (Col %) | 13 (19.4) | 9,440 (33.7) | 15,176 (52.3) | 3,506 (63.8) | 153 (75.0) | <0.0001 |
| Injury, n (Col %) ‡ | 2,461 (9.4) | 4,108 (16.3) | 917 (22.1) | 24 (17.3) | <0.0001 | |
DBS=Deep Brain Stimulation, LED=Levodopa Equivalent Dose
To avoid violating Center for Medicare and Medicaid Services cell suppression policy (no cells ≤10), these groups were combined for some characteristics
To avoid violating Center for Medicare and Medicaid Services cell suppression policy (no cells ≤10), these groups were combined for some characteristics
Out of subset sample restricted to individuals with any PD claim in 2014 (n=55,568)
Kruskal-Wallis Test
Outcomes
Table 3 presents the results of crude and adjusted logistic regression models. In total, 10.3% (n=6,471) of beneficiaries died, 27.2% (n=17,068) were hospitalized at least once, and 45.1% (n=28,288) had at least one ED visit in 2014. The odds of mortality increased with increasing frailty in both crude and adjusted models. After adjustment for sex, age category, race, residence, PD status (incident vs. prevalent), neurologist care, nursing facility residence, DBS, CCS, and mean daily LED, the odds of dying in 2014 was 1.74 times higher (95% CI: 1.62, 1.88) in the mildly frail group compared to the non-frail/prefrail group. For the moderately frail and the severely frail group, the odds were 2.38 times (95% CI: 2.13, 2.66) and 2.74 times (95% CI: 1.98, 3.78) higher than the non-frail/prefrail group, respectively.
Table 3.
Associations 2013 categorical frailty and 2014 outcomes among eligible Medicare beneficiaries with active management of Parkinson disease (PD) (n=62,786)
| 2014 Outcome | 2013 Frailty Category |
Crude 2014 Incidence per 1,000 beneficiaries |
Model 1* | Model 2* | Model 3* |
|---|---|---|---|---|---|
| OR (95%CI) | AOR (95%CI) | AOR (95%CI) | |||
| Death | |||||
| Non-Frail/Prefrail | 45 | 1.00 (REF) | 1.00 (REF) | 1.00 (REF) | |
| Mildly Frail | 124 | 2.97 (2.78, 3.17) | 2.47 (2.31, 2.64) | 1.74 (1.62, 1.88) | |
| Moderately Frail | 276 | 8.02 (7.39, 8.70) | 6.06 (5.56, 6.59) | 2.38 (2.13, 2.66) | |
| Severely Frail | 412 | 14.70 (11.06, 19.54) | 11.94 (8.89, 16.05) | 2.74 (1.98, 3.78) | |
| Hospitalization | |||||
| Non-Frail/Prefrail | 181 | 1.00 | 1.00 | 1.00 | |
| Mildly Frail | 324 | 2.17 (2.09, 2.25) | 2.05 (1.97, 2.13) | 1.69 (1.62, 1.77) | |
| Moderately Frail | 450 | 3.71 (3.49, 3.94) | 3.37 (3.17, 3.59) | 2.03 (1.88, 2.20) | |
| Severely Frail | 574 | 6.09 (4.60, 8.05) | 5.58 (4.21, 7.38) | 2.34 (1.74, 3.14) | |
| Emergency Department Visit | |||||
| Non-Frail/Prefrail | 337 | 1.00 | 1.00 | 1.00 | |
| Mildly Frail | 523 | 2.16 (2.09, 2.24) | 2.02 (1.95, 2.09) | 1.78 (1.71, 1.84) | |
| Moderately Frail | 638 | 3.47 (3.27, 3.69) | 3.09 (2.91, 3.29) | 2.26 (2.09, 2.44) | |
| Severely Frail | 750 | 5.90 (4.29, 8.11) | 5.25 (3.82, 7.23) | 2.97 (2.14, 4.13) | |
| Injury † | |||||
| Non-Frail/Prefrail | 94 | 1.00 | 1.00 | 1.00 | |
| Mildly Frail | 163 | 1.87 (1.77, 1.97) | 1.68 (1.59, 1.77) | 1.62 (1.53, 1.72) | |
| Moderately Frail | 221 | 2.72 (2.50, 2.96) | 2.25 (2.06, 2.46) | 2.09 (1.87, 2.33) | |
| Severely Frail | 173 | 2.00 (1.29, 3.12) | 1.64 (1.05, 2.56) | 1.43 (0.90, 2.26) | |
Model 1: Unadjusted
Model 2: Adjusted for sex, age category, race, and residence
Model 3: Adjusted for sex, age category, race, residence, PD status, neurologist care, nursing facility residence, deep brain stimulation, Combined Comorbidity Score, and mean daily levodopa equivalent dosage
P-value for linear trend for all analyses was <0.0001
Out of subset sample restricted to individuals with any PD claim in 2014 (n=55,568)
The crude and adjusted odds of both hospitalization and ED visit increased with more severe CFI frailty (p-value for linear trend <0.0001). After adjustment, those with severe frailty had 2.34 times greater odds of hospitalization (95% CI: 1.74, 3.14) and 2.97 times greater odds of an ED visit (95% CI: 2.14, 4.13) as compared to the non-frail/prefrail group.
In the injury subset sample, 13.5% (n=7,510) experienced an injurious fall in 2014. Compared to the non-frail/prefrail group, the fully adjusted odds of injury was 1.62 (95% CI: 1.53, 1.72) in the mildly frail and 2.09 (1.87, 2.33) in the moderately frail groups; the magnitude of increase was less and not significant in the severely frail group (OR = 1.43, 95% CI: 0.90, 2.26).
DISCUSSION
In our study of a large, national sample of older Medicare beneficiaries with PD using a previously validated CFI, we found that frailty is common, suggesting an overwhelming state of vulnerability among these individuals. Frail persons with PD were older, had more comorbidities, and had worse health outcomes than non-frail persons with PD. Finally, we found that frailty measures a phenotypic state that is beyond what is captured by PD treatment intensity alone. These findings have important implications for the care of individuals with PD and the design and analysis of observational studies of PD.
Our current findings using a CFI demonstrate that frailty is identifiable in 55.3% of Medicare beneficiaries with PD. This estimate is higher than estimates typically found in general older adult samples drawn from either community settings3,44 or Medicare beneficiary data.5 PD itself contributes to the frailty phenotype, as expected, based on studies which find that compared to controls, frailty is more common in PD.14-16,21 Previous studies of persons with PD report a wide range of frailty estimates9-23 with the highest estimate (84%) stemming from an inpatient sample.14 Our high prevalence could reflect an older sample with more severe, treated disease. However, studies with similar prevalence estimates among those with PD have been conducted in an older, community-dwelling sample (50% frail)9,20 and a younger, tertiary care center sample (52%).23 Estimates of frailty may also differ based on the instrument used and its corresponding thresholds.34 Consequently, direct comparisons of our data with prior estimates are challenging because prior studies primarily have used phenotype-centered instruments,9 rather than the CFI we utilized in our study. One prior study found an 81% agreement between a deficits-based and phenotype-based frailty instrument in frailty classification of individuals with neurodegenerative diseases.24
In our study, frailty among beneficiaries with PD was more prevalent among those that are describable as female, Black, and older. These associations are consistent with the general, older adult population;2,3,16,44 however, there is limited evidence of demographic patterns in PD frailty.10,15,17,21-23 In the general older adult population, frailty has also been associated with socioeconomic status and education level;2,16 however, we did not have individual-level data on these factors in this dataset.
After accounting for sociodemographic and clinical characteristics (including comorbidity burden), frailty was positively associated with one-year mortality, hospitalization, and ED visits among persons with PD. These findings provide evidence for the predictive validity of frailty in PD. General population studies have found similar relationships between frailty and these outcomes.5,7,33 Prior data on the longitudinal relationships between frailty and outcomes in PD are limited to a single hospital-based PD study, which found that frailty is associated with inpatient mortality, but not 30-day post-discharge mortality, institutionalization, a longer length of stay, or 30-day readmission.14
In our assessment of fall-related injuries in PD, the odds of an injury initially increased with frailty but then declined in the severely frail group. This pattern may be the result of increased mortality in the severely frail group. Alternatively, the frailest PD patients likely become less likely to fall as compared to those who are moderately frail, because they no longer attempt independent ambulation (even with assistance or assist device) and are receiving full external support and supervision for transfers or ambulation.
Several studies have found that compared to non-frail, frail individuals with PD had higher Unified Parkinson Disease Rating Scale (UPDRS) scores,15,17,19,21,22 Hoehn and Yahr Scores,10,17,21,23 and dopaminergic medication dosages.11,12,17,21,23 By directly examining the relationship between dopaminergic medication dose and frailty levels, we found that the association between frailty severity and PD medication dosage is not consistently linear, with potential clinical and research implications. Our finding is consistent with the clinical observation that PD medication efficacy and tolerance decline in advanced PD stages as patients develop cognitive impairment and increasingly convert to a dopaminergic drug-resistant postural instability and gait disorder phenotype.45 The research implications of our findings are that LED measurements may not fully account for the clinical picture that drives common outcomes like death, injurious falls, and acute care use in persons with PD.
As more PD research involves "big data" from health systems, claims datasets and disease registries, researchers need additional tools to reduce misclassification bias, particularly as an individual or cohort progresses beyond the mild/moderate disease phase. Given the need for additional safety studies of many medications for non-motor symptoms in PD,46 accounting for frailty may also be crucial to obtain unbiased effect estimates in pharmacoepidemiology studies by reducing residual confounding. Frailty may be an important confounder in PD studies more broadly.47 Finally, frailty may allow for matching persons with PD to controls in other studies, including biomarker and outcomes studies. Assessing frailty among persons with PD in a clinical setting may inform prognostic counseling and decision-making, such as palliative care referral or the need for deprescribing.14,48 Additionally, screening for frailty could result in referrals for physical therapy, exercise, and nutritional consultations to mitigate frailty.49
Our study has several strengths, including its use of a previously validated frailty instrument and a study design intended to provide inclusive, Medicare beneficiary-representative data on PD. This type of analysis could be applied to other large datasets including health system data and disease registries. Concerning limitations, claims data do not contain disease-specific measures or patient-reported outcomes, such as cognitive testing, PD motor symptom severity scores, or quality of life. Even with the deficit-based frailty approach, there is an overlap between frailty items and PD, thus we were unable to disentangle frailty from PD-related motor and non-motor symptoms.48 The lack of traditional PD severity measures in our data further limited our ability to assess the independent contribution of frailty on outcomes. However, frailty, irrespective of etiology, still has important health implications.24 LED doses may have been over- or under-estimated due to unavailable inpatient prescription data, early or late refills, titrations, and weaning.
As with all claims, registry, or observational studies, miscoding or under/over coding may affect our data and subsequent observations in a random or non-random manner. We were also unable to capture claims not in the Carrier file, such as those from durable medical equipment suppliers. Approaches to address these limitations include future studies that attempt to validate the CFI against clinically assessed frailty in PD and clinical measures of PD symptoms and severity, using data from a large, phenotyped cohort linked to claims data. Future studies may also seek to develop a claims-based index to assess PD severity, anchored to the UPDRS and other standard clinical research measures.
CONCLUSIONS
Frailty among individuals with PD can be measured in claims data. Accounting for frailty may reduce bias and improve PD observational study data analyses' methodological quality, particularly studies that use "big data" or assess drug safety and effectiveness. Given the high prevalence of frailty in PD, and its impact on outcomes, frailty assessments may have future implications for PD research and care.
Acknowledgments
Financial Disclosures/Conflicts of Interest Related to Research Covered in this Article:
Danielle S. Abraham receives research support from NIH grant #R01 NS099129 and other compensation from the Parkinson's Foundation.
Thanh Phuong Pham Nguyen receives research support from NIH grant #R01 NS099129.
Allison W. Willis receives financial support from NIH grant #R01 NS099129, the Parkinson's Foundation, the University of Pennsylvania, Biogen, Inc., and Acadia, Inc..
Study Funded by a grant from the National Institute of Neurological Disorders and Stroke of the National Institutes of Health [R01 NS099129].
REFERENCES
- 1.Campbell AJ, Buchner DM. Unstable disability and the fluctuations of frailty. Age Ageing. England; 1997;26:315–318. [DOI] [PubMed] [Google Scholar]
- 2.Fried LP, Tangen CM, Walston J, et al. Frailty in Older Adults: Evidence for a Phenotype. The Journals of Gerontology: Series A. 2001;56:M146–M157. [DOI] [PubMed] [Google Scholar]
- 3.O’Caoimh R, Sezgin D, O’Donovan MR, et al. Prevalence of frailty in 62 countries across the world: a systematic review and meta-analysis of population-level studies. Age and Ageing. 2021;50:96–104. [DOI] [PubMed] [Google Scholar]
- 4.Kim DH, Schneeweiss S. Measuring frailty using claims data for pharmacoepidemiologic studies of mortality in older adults: evidence and recommendations. Pharmacoepidemiol Drug Saf. 2014;23:891–901. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kim DH, Glynn RJ, Avorn J, et al. Validation of a Claims-Based Frailty Index Against Physical Performance and Adverse Health Outcomes in the Health and Retirement Study. J Gerontol A Biol Sci Med Sci. 2019;74:1271–1276. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Theou O, Squires E, Mallery K, et al. What do we know about frailty in the acute care setting? A scoping review. BMC Geriatrics. 2018;18:139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Vermeiren S, Vella-Azzopardi R, Beckwée D, et al. Frailty and the Prediction of Negative Health Outcomes: A Meta-Analysis. J Am Med Dir Assoc. United States; 2016;17:1163.e1–1163.e17. [DOI] [PubMed] [Google Scholar]
- 8.Kim DH. Measuring Frailty in Health Care Databases for Clinical Care and Research. Ann Geriatr Med Res. 2020/April/03 ed. Korean Geriatrics Society; 2020;24:62–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Smith N, Brennan L, Gaunt DM, Ben-Shlomo Y, Henderson E. Frailty in Parkinson’s Disease: A Systematic Review. J Parkinsons Dis. 2019;9:517–524. [DOI] [PubMed] [Google Scholar]
- 10.Roland KP, Jakobi JM, Jones GR, Powell C. Quality of life as a determinant of frailty phenotype in community-dwelling persons with Parkinson’s disease. J Am Geriatr Soc. 2012;60:590–592. [DOI] [PubMed] [Google Scholar]
- 11.Roland KP, Cornett KMD, Theou O, Jakobi JM, Jones GR. Physical Activity across Frailty Phenotypes in Females with Parkinson’s Disease. J Aging Res. 2012;2012:468156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Roland KP, Cornett KMD, Theou O, Jakobi JM, Jones GR. Concurrence of Frailty and Parkinson’s Disease. J Frailty Aging. 2012;1:123–127. [DOI] [PubMed] [Google Scholar]
- 13.Roland KP, Jones GR, Jakobi JM. Daily electromyography in females with Parkinson’s disease: a potential indicator of frailty. Arch Gerontol Geriatr. 2014;58:80–87. [DOI] [PubMed] [Google Scholar]
- 14.Torsney KM, Romero-Ortuno R. The Clinical Frailty Scale predicts inpatient mortality in older hospitalised patients with idiopathic Parkinson’s disease. J R Coll Physicians Edinb. 2018;48:103–107. [DOI] [PubMed] [Google Scholar]
- 15.Tan AH, Hew YC, Lim S-Y, et al. Altered body composition, sarcopenia, frailty, and their clinico-biological correlates, in Parkinson’s disease. Parkinsonism Relat Disord. 2018;56:58–64. [DOI] [PubMed] [Google Scholar]
- 16.Liotta G, O’Caoimh R, Gilardi F, et al. Assessment of frailty in community-dwelling older adults residents in the Lazio region (Italy): A model to plan regional community-based services. Arch Gerontol Geriatr. 2017;68:1–7. [DOI] [PubMed] [Google Scholar]
- 17.Lin W-C, Huang Y-C, Leong C-P, et al. Associations Between Cognitive Functions and Physical Frailty in Patients With Parkinson’s Disease. Front Aging Neurosci. 2019;11:283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Buchman AS, Yu L, Wilson RS, Schneider JA, Bennett DA. Association of brain pathology with the progression of frailty in older adults. Neurology. 2013;80:2055–2061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ahmed NN, Sherman SJ, Vanwyck D. Frailty in Parkinson’s disease and its clinical implications. Parkinsonism Relat Disord. 2008;14:334–337. [DOI] [PubMed] [Google Scholar]
- 20.Renne I, Gobbens RJ. Effects of frailty and chronic diseases on quality of life in Dutch community-dwelling older adults: a cross-sectional study. Clin Interv Aging. 2018;13:325–334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Peball M, Mahlknecht P, Werkmann M, et al. Prevalence and Associated Factors of Sarcopenia and Frailty in Parkinson’s Disease: A Cross-Sectional Study. Gerontology. Switzerland; 2019;65:216–228. [DOI] [PubMed] [Google Scholar]
- 22.Smith N, Gaunt D, Whone A, Ben-Shlomo Y, Henderson E. The association between frailty and Parkinson’s in the ReSPOnD trial [abstract]. Movement Disorders [online serial]. 2019;34. Accessed at: https://www.mdsabstracts.org/abstract/the-association-between-frailty-and-parkinsons-in-the-respond-trial/. Accessed December 9, 2020. [Google Scholar]
- 23.Firat Özer F, Akin S, Gültekin M, Ertürk Zararsiz G, Ekinci Soylu A. Frailty in Patients with Parkinson’s Disease: Associations with Disability and Timed Up and Go. Arch Neuropsychiatry. Epub 2019.:1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Burt JR, Godin J, Filion J, et al. Frailty Prevalence in the COMPASS-ND Study of Neurodegenerative Disorders. Can Geriatr J. Canadian Geriatrics Society; 2019;22:205–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Willis AW, Schootman M, Evanoff BA, Perlmutter JS, Racette BA. Neurologist care in Parkinson disease: a utilization, outcomes, and survival study. Neurology. 2011;77:851–857. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Bloem BR, Ypinga JHL, Willis A, et al. Using Medical Claims Analyses to Understand Interventions for Parkinson Patients. J Parkinsons Dis. 2018;8:45–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Willis AW. Using Administrative Data to Examine Health Disparities and Outcomes in Neurological Diseases of the Elderly. Current Neurology and Neuroscience Reports. 2015;15:75. [DOI] [PubMed] [Google Scholar]
- 28.Buta BJ, Walston JD, Godino JG, et al. Frailty assessment instruments: Systematic characterization of the uses and contexts of highly-cited instruments. Ageing Res Rev. 2016;26:53–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Rockwood K, Mitnitski A. Frailty in Relation to the Accumulation of Deficits. The Journals of Gerontology: Series A. 2007;62:722–727. [DOI] [PubMed] [Google Scholar]
- 30.Research Data Assistance Center (ResDAC). Research Identifiable File Availability [online]. 2019January. Accessed at: https://www.resdac.org/file-availability.Accessed January 22, 2019.
- 31.Chronic Conditions Data Warehouse. Condition Categories [online]. Centers for Medicare & Medicaid Services; 2019. Accessed at: https://www.ccwdata.org/web/guest/condition-categories. Accessed October 18, 2020. [Google Scholar]
- 32.Butt DA, Tu K, Young J, et al. A validation study of administrative data algorithms to identify patients with Parkinsonism with prevalence and incidence trends. Neuroepidemiology. Switzerland; 2014;43:28–37. [DOI] [PubMed] [Google Scholar]
- 33.Kim DH, Schneeweiss S, Glynn RJ, Lipsitz LA, Rockwood K, Avorn J. Measuring Frailty in Medicare Data: Development and Validation of a Claims-Based Frailty Index. J Gerontol A Biol Sci Med Sci. 2018;73:980–987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Kim DH, Patorno E, Pawar A, Lee H, Schneeweiss S, Glynn RJ. Measuring Frailty in Administrative Claims Data: Comparative Performance of Four Claims-Based Frailty Measures in the U.S. Medicare Data. The Journals of Gerontology: Series A. 2020;75:1120–1125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Kim DH, Gautam N. Research Background - Claims-Based Frailty Index 20200903.docx [online]. V6 ed. SAS Programs - Claims-Based Frailty Index Harvard Dataverse; 2020. Accessed at: 10.7910/DVN/HM8DOI/HVLESQ. [DOI] [Google Scholar]
- 36.Kim DH, Gautam N. SAS Programs - Claims-Based Frailty Index [online]. V9 ed. Harvard Dataverse; 2020. Accessed at: 10.7910/DVN/HM8DOI. [DOI] [Google Scholar]
- 37.Caldwell D, Merriman K, Durham S. How to Identify Hospital Claims for Emergency Room Visits in the Medicare Claims Data [online]. ResDAC; 2015. July. Accessed at: https://www.resdac.org/articles/how-identify-hospital-claims-emergency-room-visits-medicare-claims-data. Accessed September 17, 2020. [Google Scholar]
- 38.Hoffman GJ, Hays RD, Shapiro MF, Wallace SP, Ettner SL. Claims-based Identification Methods and the Cost of Fall-related Injuries Among US Older Adults. Med Care. 2016;54:664–671. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Economic Research Service. Rural-Urban Continuum Codes [online]. United States Department of Agriculture; 2016. Accessed at: https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/. Accessed September 14, 2020. [Google Scholar]
- 40.Koroukian SM, Xu F, Murray P. Ability of Medicare claims data to identify nursing home patients: a validation study. Med Care. 2008;46:1184–1187. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Centers for Medicare & Medicaid Services. Nursing Facility Services (Codes 99304 - 99318) [online]. 2006January. Report No.: MM4246. Accessed at: https://www.cms.gov/Outreach-and-Education/Medicare-Learning-Network-MLN/MLNMattersArticles/downloads/MM4246.pdf.Accessed October 18, 2020.
- 42.Gagne JJ, Glynn RJ, Avorn J, Levin R, Schneeweiss S. A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol. 2011;64:749–759. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Tomlinson CL, Stowe R, Patel S, Rick C, Gray R, Clarke CE. Systematic review of levodopa dose equivalency reporting in Parkinson’s disease. Mov Disord. 2010;25:2649–2653. [DOI] [PubMed] [Google Scholar]
- 44.Collard RM, Boter H, Schoevers RA, Oude Voshaar RC. Prevalence of frailty in community-dwelling older persons: a systematic review. J Am Geriatr Soc. United States; 2012;60:1487–1492. [DOI] [PubMed] [Google Scholar]
- 45.Lizarraga KJ, Fox SH, Strafella AP, Lang AE. Hallucinations, Delusions and Impulse Control Disorders in Parkinson Disease. Clin Geriatr Med. United States; 2020;36:105–118. [DOI] [PubMed] [Google Scholar]
- 46.Seppi K, Ray Chaudhuri K, Coelho M, et al. Update on treatments for nonmotor symptoms of Parkinson’s disease-an evidence-based medicine review. Mov Disord. 2019;34:180–198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Tan AH, Chong CW, Lim S-Y, et al. Gut Microbial Ecosystem in Parkinson Disease: New Clinicobiological Insights from Multi-Omics. Ann Neurol; 2021;89:546–559.. [DOI] [PubMed] [Google Scholar]
- 48.Tenison E, Henderson EJ. Multimorbidity and Frailty: Tackling Complexity in Parkinson’s Disease. J Parkinsons Dis. IOS Press; 2020;10:S85–S91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Puts MTE, Toubasi S, Andrew MK, et al. Interventions to prevent or reduce the level of frailty in community-dwelling older adults: a scoping review of the literature and international policies. Age Ageing. 2017;46:383–392. [DOI] [PMC free article] [PubMed] [Google Scholar]