Disentangling the Disabling Process
Disentangling the Disabling Process
The assembly of the cohort is summarized in Figure 2 and has been described in detail elsewhere (Gill, Desai, Gahbauer, Holford, & Williams, 2001; Hardy & Gill, 2004). Potential participants were identified from a computerized list of 3,157 age-eligible members of a large health plan in greater New Haven, Connecticut. Eligibility was determined during a screening telephone interview and was confirmed during an in-home assessment. Members were potentially eligible if they were nondisabled (i.e., required no personal assistance) in four basic ADLs—bathing, dressing, walking inside the house, and transferring from a chair. Exclusion criteria included significant cognitive impairment with no available proxy, inability to speak English, diagnosis of a terminal illness, and a plan to move out of the area during the next year.
(Enlarge Image)
Figure 2.
Assembly of the study cohort. Persons who were physically frail were oversampled. After the prespecified number of nonfrail participants were enrolled, potential participants were excluded if they had a low likelihood of physical frailty based on the telephone screen and, subsequently, if they were found not to be physically frail during the in-home assessment.
Based on our initial sample size calculations, persons were oversampled if they were physically frail, as denoted by a timed score greater than 10 s on the rapid gait test (i.e., walk back and forth over a 10-ft [3-m] course as quickly as possible; (Gill, Williams, & Tinetti, 1995). In the absence of a gold standard, operationalizing physical frailty as slow gait speed was justified by its high face validity (Goodwin, 2002), clinical feasibility (Gill, McGloin, Gahbauer, Shepard, & Bianco, 2001), and strong epidemiologic link to functional decline and disability (Abellan van Kan et al., 2009; Gill et al., 1995; Guralnik, 1994). Only 4.6% of the 2,753 health plan members who were alive and could be contacted refused to complete the screening telephone interview, and 75.2% of the eligible members agreed to participate and were enrolled between March 1998 and October 1999. Persons who refused to participate did not differ significantly from those who were enrolled in terms of age or sex. The study protocol was approved by the Yale Human Investigation Committee, and all participants provided verbal informed consent.
Comprehensive home-based assessments have been completed at 18-month intervals (except at 126 months), and telephone interviews have been completed monthly. For participants who have significant cognitive impairment or are otherwise unavailable, a proxy is interviewed using a rigorous protocol, with demonstrated reliability and validity (Gill, Hardy, & Williams, 2002).
Comprehensive Assessments.—The core elements, included at each time point, are provided in Table 1 . Height and other demographic information, such as education, were ascertained at baseline only. The nine self-reported, physician-diagnosed chronic conditions include hypertension, myocardial infarction, congestive heart failure, stroke, diabetes mellitus, arthritis, hip fracture, chronic lung disease, and cancer. Based on prior work, which found that questions about difficulty and dependence provide complementary information (Gill et al., 1998), questions about both difficulty and dependence have been included for each of the seven basic activities (bathing, dressing, transferring, walking, eating, toileting, and grooming), five instrumental activities (shopping, housework, meal preparation, taking medications, and managing finances), and three mobility tasks (walk a quarter mile, climb flight of stairs, and lift/carry 10 pounds). A fourth mobility item, also scored as 0, 1, or 2, was based on the average amount of time (in hours) walked per day (>0.75, 0.25–0.75, or <0.25; Gill, Allore, & Guo, 2004). To enhance the quality of the data, the need for a proxy informant is evaluated during each comprehensive assessment. From the available data, slightly modified versions of the Fried Frailty Phenotype (Gill, Gahbauer, Allore, & Han, 2006) and Short Physical Performance Battery (Gill, Murphy, Barry, & Allore, 2009) have been developed. In addition, expanded modules on bathing (Naik, Concato, & Gill, 2004; Naik & Gill, 2005), sleep (Vaz Fragoso, Gahbauer, Van Ness, & Gill, 2009), and fatigue (Yellen, Cella, Webster, Blendowski, & Kaplan, 1997) were added to the comprehensive assessments starting at 36, 90, and 108 months, respectively; and age stereotypes (Levy, Slade, & Gill, 2006; Levy, Slade, Murphy, & Gill, 2012) were assessed at baseline and 108 months. With few exceptions, data on the core elements were 100% complete at baseline and greater than 95% complete during the subsequent comprehensive assessments. To account for these missing data, we have used sequential Markov Chain Monte Carlo imputation for multivariate normal data.
Monthly Interviews.—With the support of a computer-aided telephone interview, participants are interviewed monthly to ascertain their exposure to intervening events, monitor their health care utilization, and reassess their functional status. The intervening events include illnesses and injuries leading to either hospitalization or restricted activity (Gill, Allore, Holford, & Guo, 2004b). Participants are asked whether they had stayed at least overnight in a hospital since the last interview, that is, during the past month. To ascertain less potent intervening events, participants are asked two questions related to restricted activity: (a) "Since we last talked on (date of last interview), have you cut down on your usual activities due to an illness, injury or other problem?" and (b) "Since we last talked on (date of last interview), have you stayed in bed for at least half a day due to an illness, injury or other problem?" Participants are considered to have restricted activity if they answered "Yes" to one or both of the questions. If participants have restricted activity, they are asked sequentially whether they have had any of 24 prespecified problems ( Table 2 ) "since we last talked on (date of last interview)."
Participants who have been hospitalized are asked the name of the hospital and the primary reason for their admission. These reasons are subsequently grouped into distinct diagnostic categories using a revised version of the protocol described by Ferrucci and coworkers (Ferrucci, Guralnik, Pahor, Corti, & Havlik, 1997; Gill, Allore, et al., 2004b). Participants are also asked whether they had seen a doctor in the office or an emergency room since their last interview and the primary reason for these visits. An additional set of questions asks specifically about seeing a psychiatrist, psychologist, or counselor, and about admissions to a nursing home (or hospice) since the last interview (Gill, Allore, & Han, 2006).
Finally, participants are asked, "At the present time, do you need help from another person to (complete the task)?" for each of four basic activities (bathing, dressing, walking inside the house, and transferring from a chair), five instrumental activities (shopping, housework, meal preparation, taking medications, and managing finances), and three mobility activities (walk ¼ mile, climb flight of stairs, and lift/carry 10 pounds). For these 12 activities, disability is operationalized as the need for personal assistance (Gill, Murphy, Gahbauer, & Allore, 2013a). Participants are also asked about a fourth mobility activity, "Have you driven a car during the past month?" Participants who respond "No" are deemed to have stopped driving. To maintain consistency with the other activities, these participants are classified as being "disabled" in driving that month (Gill, Gahbauer, Murphy, Han, & Allore, 2012).
In accordance with recommendations for binary longitudinal data (Wang & Fitzmaurice, 2006), we have used multiple imputation to address the small amount of missing monthly data on functional status (Gill, Guo, & Allore, 2008). Missing data have not been imputed for the intervening events.
Hospital Records.—For all self-reported hospital admissions, we obtain discharge summaries and extract information, using standard procedures (Inouye et al., 1998), on dates of admission and discharge, diagnosis on admission, primary and other (up to 16) diagnoses at discharge (with International Classification of Diseases, Ninth Revision [ICD]-9 codes), major procedures (≤12 with ICD-9 codes), expected source of payment (e.g., Medicare and Medicaid), and discharge location (e.g., home and nursing home). Although our ultimate goal is to obtain and review such records for all admissions, for efficiency in the context of an ongoing longitudinal study, we have focused our efforts to date primarily on decedents because they can have no additional hospital admissions.
Deaths. Deaths are ascertained from local obituaries and/or an informant during a subsequent interview. From the informant, we obtain information on date of death, hospitalization since the last interview, reason for the hospitalization, nursing home (or hospice) admission since the last interview, site and cause of death, whether the death was expected versus sudden/unexpected, and the need for (and duration of) personal assistance with the 12 basic, instrumental, and mobility activities during the last year of life. In addition, we obtain a copy of the death certificate and have a certified nosologist provide us with the ICD-10 codes for the immediate and underlying causes of death (Gill, Gahbauer, Han, & Allore, 2010).
Medicare Data.—In 2011, we obtained detailed participant-level data on health care utilization (from 1997 forward) through linkages with Medicare claims, using procedures adapted from prior studies (Wolinsky et al., 2007). These claims are based on the information needed to process and pay bills for persons insured by Medicare. The Medicare denominator file (1998), Beneficiary Summary File (up to 2009), and Master Beneficiary Summary File (after 2009) contain monthly managed care indicators (yes/no), denoting whether the beneficiary is in Medicare Fee-For-Service (FFS) or managed care (Part C). Monthly entitlement indicators, denoting Parts A, B, and D, are also provided. Claims are divided into files based on billing form and location of care (inpatient hospital, outpatient, skilled nursing facility, hospice, home health; Centers for Medicare & Medicaid Services, 2014). More recently, we have obtained files on durable medical equipment and assessment data from the Long-Term Care Minimum Data Set (MDS), available for all participants who are in Medicare or Medicaid nursing facilities (Rahman & Applebaum, 2009) and from the Home Health Outcome and Assessment Information Set, available for participants receiving Medicare-supported home care services (Fortinsky, Garcia, Joseph Sheehan, Madigan, & Tullai-McGuinness, 2003). Updates of these files are obtained annually.
A successful match to Medicare claims has been made for all but one of the 754 participants. Because the participants were originally members of a large health plan, nearly half (49.9%) were in managed Medicare during at least part of the follow-up period, with an overall mean (SD) penetrance (per 100 person-month) of 23.0 (12.3). This value has ranged from a high of 46.0 in 1999 to a low of 12.1 in 2004. MDS assessment data and claims for hospice care are included in the files regardless of plan type, that, FES or managed care.
Precipitating Events Project
Study Population
The assembly of the cohort is summarized in Figure 2 and has been described in detail elsewhere (Gill, Desai, Gahbauer, Holford, & Williams, 2001; Hardy & Gill, 2004). Potential participants were identified from a computerized list of 3,157 age-eligible members of a large health plan in greater New Haven, Connecticut. Eligibility was determined during a screening telephone interview and was confirmed during an in-home assessment. Members were potentially eligible if they were nondisabled (i.e., required no personal assistance) in four basic ADLs—bathing, dressing, walking inside the house, and transferring from a chair. Exclusion criteria included significant cognitive impairment with no available proxy, inability to speak English, diagnosis of a terminal illness, and a plan to move out of the area during the next year.
(Enlarge Image)
Figure 2.
Assembly of the study cohort. Persons who were physically frail were oversampled. After the prespecified number of nonfrail participants were enrolled, potential participants were excluded if they had a low likelihood of physical frailty based on the telephone screen and, subsequently, if they were found not to be physically frail during the in-home assessment.
Based on our initial sample size calculations, persons were oversampled if they were physically frail, as denoted by a timed score greater than 10 s on the rapid gait test (i.e., walk back and forth over a 10-ft [3-m] course as quickly as possible; (Gill, Williams, & Tinetti, 1995). In the absence of a gold standard, operationalizing physical frailty as slow gait speed was justified by its high face validity (Goodwin, 2002), clinical feasibility (Gill, McGloin, Gahbauer, Shepard, & Bianco, 2001), and strong epidemiologic link to functional decline and disability (Abellan van Kan et al., 2009; Gill et al., 1995; Guralnik, 1994). Only 4.6% of the 2,753 health plan members who were alive and could be contacted refused to complete the screening telephone interview, and 75.2% of the eligible members agreed to participate and were enrolled between March 1998 and October 1999. Persons who refused to participate did not differ significantly from those who were enrolled in terms of age or sex. The study protocol was approved by the Yale Human Investigation Committee, and all participants provided verbal informed consent.
Data Collection
Comprehensive home-based assessments have been completed at 18-month intervals (except at 126 months), and telephone interviews have been completed monthly. For participants who have significant cognitive impairment or are otherwise unavailable, a proxy is interviewed using a rigorous protocol, with demonstrated reliability and validity (Gill, Hardy, & Williams, 2002).
Comprehensive Assessments.—The core elements, included at each time point, are provided in Table 1 . Height and other demographic information, such as education, were ascertained at baseline only. The nine self-reported, physician-diagnosed chronic conditions include hypertension, myocardial infarction, congestive heart failure, stroke, diabetes mellitus, arthritis, hip fracture, chronic lung disease, and cancer. Based on prior work, which found that questions about difficulty and dependence provide complementary information (Gill et al., 1998), questions about both difficulty and dependence have been included for each of the seven basic activities (bathing, dressing, transferring, walking, eating, toileting, and grooming), five instrumental activities (shopping, housework, meal preparation, taking medications, and managing finances), and three mobility tasks (walk a quarter mile, climb flight of stairs, and lift/carry 10 pounds). A fourth mobility item, also scored as 0, 1, or 2, was based on the average amount of time (in hours) walked per day (>0.75, 0.25–0.75, or <0.25; Gill, Allore, & Guo, 2004). To enhance the quality of the data, the need for a proxy informant is evaluated during each comprehensive assessment. From the available data, slightly modified versions of the Fried Frailty Phenotype (Gill, Gahbauer, Allore, & Han, 2006) and Short Physical Performance Battery (Gill, Murphy, Barry, & Allore, 2009) have been developed. In addition, expanded modules on bathing (Naik, Concato, & Gill, 2004; Naik & Gill, 2005), sleep (Vaz Fragoso, Gahbauer, Van Ness, & Gill, 2009), and fatigue (Yellen, Cella, Webster, Blendowski, & Kaplan, 1997) were added to the comprehensive assessments starting at 36, 90, and 108 months, respectively; and age stereotypes (Levy, Slade, & Gill, 2006; Levy, Slade, Murphy, & Gill, 2012) were assessed at baseline and 108 months. With few exceptions, data on the core elements were 100% complete at baseline and greater than 95% complete during the subsequent comprehensive assessments. To account for these missing data, we have used sequential Markov Chain Monte Carlo imputation for multivariate normal data.
Monthly Interviews.—With the support of a computer-aided telephone interview, participants are interviewed monthly to ascertain their exposure to intervening events, monitor their health care utilization, and reassess their functional status. The intervening events include illnesses and injuries leading to either hospitalization or restricted activity (Gill, Allore, Holford, & Guo, 2004b). Participants are asked whether they had stayed at least overnight in a hospital since the last interview, that is, during the past month. To ascertain less potent intervening events, participants are asked two questions related to restricted activity: (a) "Since we last talked on (date of last interview), have you cut down on your usual activities due to an illness, injury or other problem?" and (b) "Since we last talked on (date of last interview), have you stayed in bed for at least half a day due to an illness, injury or other problem?" Participants are considered to have restricted activity if they answered "Yes" to one or both of the questions. If participants have restricted activity, they are asked sequentially whether they have had any of 24 prespecified problems ( Table 2 ) "since we last talked on (date of last interview)."
Participants who have been hospitalized are asked the name of the hospital and the primary reason for their admission. These reasons are subsequently grouped into distinct diagnostic categories using a revised version of the protocol described by Ferrucci and coworkers (Ferrucci, Guralnik, Pahor, Corti, & Havlik, 1997; Gill, Allore, et al., 2004b). Participants are also asked whether they had seen a doctor in the office or an emergency room since their last interview and the primary reason for these visits. An additional set of questions asks specifically about seeing a psychiatrist, psychologist, or counselor, and about admissions to a nursing home (or hospice) since the last interview (Gill, Allore, & Han, 2006).
Finally, participants are asked, "At the present time, do you need help from another person to (complete the task)?" for each of four basic activities (bathing, dressing, walking inside the house, and transferring from a chair), five instrumental activities (shopping, housework, meal preparation, taking medications, and managing finances), and three mobility activities (walk ¼ mile, climb flight of stairs, and lift/carry 10 pounds). For these 12 activities, disability is operationalized as the need for personal assistance (Gill, Murphy, Gahbauer, & Allore, 2013a). Participants are also asked about a fourth mobility activity, "Have you driven a car during the past month?" Participants who respond "No" are deemed to have stopped driving. To maintain consistency with the other activities, these participants are classified as being "disabled" in driving that month (Gill, Gahbauer, Murphy, Han, & Allore, 2012).
In accordance with recommendations for binary longitudinal data (Wang & Fitzmaurice, 2006), we have used multiple imputation to address the small amount of missing monthly data on functional status (Gill, Guo, & Allore, 2008). Missing data have not been imputed for the intervening events.
Hospital Records.—For all self-reported hospital admissions, we obtain discharge summaries and extract information, using standard procedures (Inouye et al., 1998), on dates of admission and discharge, diagnosis on admission, primary and other (up to 16) diagnoses at discharge (with International Classification of Diseases, Ninth Revision [ICD]-9 codes), major procedures (≤12 with ICD-9 codes), expected source of payment (e.g., Medicare and Medicaid), and discharge location (e.g., home and nursing home). Although our ultimate goal is to obtain and review such records for all admissions, for efficiency in the context of an ongoing longitudinal study, we have focused our efforts to date primarily on decedents because they can have no additional hospital admissions.
Deaths. Deaths are ascertained from local obituaries and/or an informant during a subsequent interview. From the informant, we obtain information on date of death, hospitalization since the last interview, reason for the hospitalization, nursing home (or hospice) admission since the last interview, site and cause of death, whether the death was expected versus sudden/unexpected, and the need for (and duration of) personal assistance with the 12 basic, instrumental, and mobility activities during the last year of life. In addition, we obtain a copy of the death certificate and have a certified nosologist provide us with the ICD-10 codes for the immediate and underlying causes of death (Gill, Gahbauer, Han, & Allore, 2010).
Medicare Data.—In 2011, we obtained detailed participant-level data on health care utilization (from 1997 forward) through linkages with Medicare claims, using procedures adapted from prior studies (Wolinsky et al., 2007). These claims are based on the information needed to process and pay bills for persons insured by Medicare. The Medicare denominator file (1998), Beneficiary Summary File (up to 2009), and Master Beneficiary Summary File (after 2009) contain monthly managed care indicators (yes/no), denoting whether the beneficiary is in Medicare Fee-For-Service (FFS) or managed care (Part C). Monthly entitlement indicators, denoting Parts A, B, and D, are also provided. Claims are divided into files based on billing form and location of care (inpatient hospital, outpatient, skilled nursing facility, hospice, home health; Centers for Medicare & Medicaid Services, 2014). More recently, we have obtained files on durable medical equipment and assessment data from the Long-Term Care Minimum Data Set (MDS), available for all participants who are in Medicare or Medicaid nursing facilities (Rahman & Applebaum, 2009) and from the Home Health Outcome and Assessment Information Set, available for participants receiving Medicare-supported home care services (Fortinsky, Garcia, Joseph Sheehan, Madigan, & Tullai-McGuinness, 2003). Updates of these files are obtained annually.
A successful match to Medicare claims has been made for all but one of the 754 participants. Because the participants were originally members of a large health plan, nearly half (49.9%) were in managed Medicare during at least part of the follow-up period, with an overall mean (SD) penetrance (per 100 person-month) of 23.0 (12.3). This value has ranged from a high of 46.0 in 1999 to a low of 12.1 in 2004. MDS assessment data and claims for hospice care are included in the files regardless of plan type, that, FES or managed care.
