Exploring geriatric assessment-driven rehabilitation referral patterns and its influence on functional outcomes and survival in older adults with advanced cancer

Abstract

Background:

Older adults with advanced cancer experience functional disability that warrants rehabilitation services; however, evidence indicates inconsistencies in referral. The purpose was to (1) identify predictors of geriatric assessment (GA)-driven referrals to rehabilitation services and (2) explore associations between referral and change in function, health-related quality of life (HRQoL), and overall survival among older adults with advanced cancer.

Methods:

This was a secondary analysis (NCT020107443, UG1CA189961) of a nationwide GA clinical trial. Patients were older adults with advanced cancer who had at least one GA-defined physical performance or functional status impairment. Primary outcomes were oncologist-initiated discussion about or referral to rehabilitation services after the GA (Aim 1) and decline in activities of daily living (ADL), Instrumental ADL (IADL), and HRQoL within 3 months, and overall survival at 1 year (Exploratory Aims). Analyses included multivariable logistic regression and Cox proportional hazards models. Demographic and clinical factors were controlled for by using 1:1 propensity score matching.

Results:

In total 265 patients were analyzed. After adjustment, impaired cognition (odds ratio [OR] = 2.25, p = 0.01), Karnofsky score indicating disability (OR = 2.86, p < 0.01), and receipt of monoclonal antibodies (OR = 1.95, p = 0.04) were associated with higher odds of referral. In contrast, polypharmacy was associated with lower odds of referral (OR = 0.31, p < 0.01). Referred patients were less likely to decline in ADL (OR 0.30, p = 0.07) and IADL (OR 0.64, p = 0.35), but more likely to decline in HRQoL (OR 1.20, p = 0.67) and have worse survival (HR 1.18, p = 0.62).

Conclusions:

Cancer treatment, polypharmacy, cognition, and disability status likely influence oncologists’ decision to refer for rehabilitation. Referral was not independently associated with change in functional disability, HRQoL, or survival. Future studies should evaluate patients’ utilization of rehabilitation services post-referral and determine whether dose/timing of rehabilitation services influence clinical outcomes.

INTRODUCTION

Due to compounding effects of aging, cancer, and treatment, older adults with advanced cancer are vulnerable to declines in physical performance, activities of daily living (ADL), and instrumental activities of daily living (IADL).¹ Nearly two-thirds of older adults with advanced cancer report clinically significant physical performance and functional deficits,² which are associated with unanticipated hospitalizations,³ chemotherapy toxicity,⁴ shorter survival,⁵ and poorer health-related quality of life (HRQoL).³ Evaluation and treatment by physical or occupational therapy practitioners (i.e., rehabilitation services) are reimbursable services that may prevent, attenuate, or treat these deficits in individuals with cancer.⁶ Inconsistent screening for physical and functional deficits, and non-standardized referral practices have led to inadequate access to rehabilitation services in this population.^7,8

Geriatric assessment (GA) is a collection of validated screening tools measuring aging-related concerns (physical performance, functional status, comorbidity, cognition, nutrition, social support, polypharmacy, and psychological status) among older adults with advanced cancer,^9,10 and guides referral recommendations, including to rehabilitation services, to address concerns.¹¹ Despite its utility, evidence synthesizing GA-driven referral recommendations on clinical outcomes is still developing. Emerging research suggests that patient characteristics (e.g., demographic, socioeconomic, and clinical factors) may impact providers’ treatment decisions and, ultimately, care quality.¹² As such, patient-level factors associated with referral to rehabilitation services may also be associated with survivorship outcomes.¹⁰ Thus, there is premise to uncover the relationship between GA-driven referral to rehabilitation services with clinical outcomes in this population.^6,8,13

The purpose of this study was to identify patient-level factors of oncologist-initiated discussions about and/or referral to rehabilitation services among older adults with advanced cancer who had physical and functional deficits identified by GA. Exploratory analyses examined the association between provider-implemented rehabilitation referrals with change in function, HRQoL, and survival in patients who received a referral compared with those who were not referred but had indications for services.

METHODS

Study Design

This was a secondary analysis of the Improving Communication in Older Cancer Patients and Their Caregivers (COACH) trial (Trial Registration: NCT020107443; UG1CA189961), conducted at 31 community oncology practices within the University of Rochester Cancer Center National Cancer Institute (NCI) Community Oncology Research Program.¹⁴ Patients were older adults (≥70 years) with an advanced solid tumor or lymphoma considering or receiving cancer treatment with palliative intent who had at least one GA-identified impairment. Each practice received approval from their respective institutional review boards, and all patients provided written informed consent. The primary objective of the parent trial was to determine whether provision of a GA summary and subsequent GA-driven recommendations to the oncology provider improved patient-provider communication about aging-related concerns. Full details of this trial are previously published.¹⁴ Briefly, enrolled patients received a GA evaluation administered by research coordinators. GA summary and recommendations were then provided to the patient’s oncologists in the intervention arm. At study entry, oncologists received a brief training about GA and had autonomy as to how they chose to use GA with patients.¹⁴ Oncologists in the usual care arm did not receive training, GA summary, or recommendations.

Participants and procedures

The study sample included individuals who were (1) enrolled in COACH intervention arm only; (2) completed a baseline GA; and (3) had a physical or functional deficit identified by the baseline GA (Figure 1). One or both impairments were required for inclusion to elicit recommendation for rehabilitation referral. Physical deficits were based on an impairment in one or more of the following assessments: Short Physical Performance Battery (≤9 points);¹⁵ Timed Up and Go (>13.5 s);¹⁶ self-report fall within the past 6 months;¹⁷ and Medical Outcomes Study (MOS) Physical Functioning Scale (self-report of significant difficulty with any task).¹⁸ A functional deficit was based either a self-reported deficit in any ADL¹⁹ or IADL.²⁰ Each patient’s treating oncologist received GA-driven recommendations, including a recommendation to refer patients to rehabilitation services if they had a physical and/or functional deficit. The oncologist or medical staff recorded whether (1) the oncologist-initiated a verbal discussion about rehabilitation with the patient; and/or (2) provided an order for rehabilitation (henceforth, called an “implemented referral”) on a data collection form (see Supplemental Figure S1).

FIGURE 1.

Study flow chart. COACH, improving communication in older cancer patients and their caregivers; HRQol, health-related quality of life; OT, occupational therapy; PT, physical therapy; PS, propensity score.

Covariables

Demographic and clinical characteristics

Age, gender, race, ethnicity, income, marital status, living arrangement, education, living situation, and emotional distress²¹ were obtained via self-report at baseline. Cancer type, treatment use (chemotherapy, radiation, surgery, monoclonal antibodies; yes/no), and the physician-reported Karnofsky Performance Status²² (i.e., no disability versus any disability) were reported by oncologists and confirmed by medical record review.

GA domain impairments

Impairments in the remaining six GA domains (i.e., cognitive function, polypharmacy, psychological function, social support, comorbidity, and nutritional status) were considered as covariables. Impaired cognition was based on the Short Blessed Orientation-Memory-Concentration²³ (≥11) and/or Mini-Cog²⁴ (0 words recalled or 1–2 words recalled with abnormal clock drawing test). Polypharmacy status was determined by the number of regularly scheduled prescription medications (≥5), or any use of a high-risk medications, or insufficient creatinine clearance (<60 mL/min) based on clinical records. Psychological impairment was based on Generalized Anxiety Disorder-7 (≥10 points)²⁵ or the Geriatric Depression Scale short form (≥5 points).²⁶ Impaired social support was based on the MOS Social Support tangible support subscale where individuals indicated less than adequate support for care on one or more item.¹⁸ Impaired comorbidity was based on the OARS Comorbidity scale (individuals answered “yes” to ≥3 chronic illnesses or one illness interferes “a great deal” with quality of life).¹⁰ Lastly, impaired nutrition was based on under-weight body mass index (<21 kg/m²)²⁷ or Mini Nutritional Status (≤11 points).²⁷

Data analysis

The primary outcome of this study was whether a patient received an oncologist-initiated discussion about and/or implemented referral to rehabilitation services. If the patient received one or both, they were considered part of the “rehabilitation referral group.” This outcome variable does not represent confirmed utilization of rehabilitation services. Descriptive statistics summarized demographic, clinical, and GA domain-related characteristics of the analytic sample and by rehabilitation referral group. Univariable logistic regression models estimated the associations between demographic, clinical, and GA domain-related characteristics on rehabilitation referral. Multivariable logistic regression models estimated factors independently associated with rehabilitation referral. Model covariates included potential confounders associated with rehabilitation referral (determined a priori: age, gender, cancer type, social support, comorbidity),^7,28,29 and significant factors (p < 0.1) identified in the univariable models. Sensitivity analyses examined the interaction between treatment and cancer type to determine if findings related to treatment type were influenced by a specific cancer type. Variance inflation factor (≥5) and the Pearson’s correlation coefficient were examined to indicate if multicollinearity existed among independent variables. Odds ratios (ORs) with 95% confidence intervals (CIs) were reported. All statistical analyses were performed using SAS version 9.4 statistical software (SAS Institute, Cary, NC).

Exploratory analyses

Exploratory analyses analyzed the association of implemented referrals with (1) decline in functional status (ADL¹⁹ and IADL²⁰) and HRQoL (Functional Assessment of Cancer Therapy—General (FACT-G)³⁰) at 3 months after GA, and (2) 1-year survival. For these analyses, we compared (1) patients that had received an implemented referral from their oncologist to (2) patients who did not receive either a discussion about or an implemented referral to rehabilitation (Figure 1). Patients who had only a discussion about referral were excluded, because an implemented referral is necessary to engage in rehabilitation services.^31,32 Sensitivity analyses were performed to examine descriptive differences in those who had a discussion only compared with those with an implemented referral (i.e., the patient received a completed order or referral to rehabilitation services; this does not consider utilization). This approach provided additional homogeneity of the sample to draw conclusions about the association between referral and exploratory outcomes. Patients were censored if they survived ≥365 days or were lost to follow-up. For all functional status and HRQoL, we considered ≥1 unit decline to indicate clinically significant change.^30,33,34

Oncologists may be more likely to recommend rehabilitation to patients who are frail and more likely to experience functional decline, leading to a potential for bias by indication. To minimize bias, patients with an implemented referral were matched 1:1 to non-referred patients based on propensity scores using established methods and recommended reporting guidelines.^35,36 Comprehensive details on propensity score creation are in Supplementary Tables S1 and S2.

After assembly of the matched sample, conditional logistic regression modeling, with a strata statement accounting for matched pairs, estimated the relative odds of decline in ADL, IADL, and HRQoL at 3 months. Patients missing data at 3 months had their 4–6 week follow-up values carried forward. Cox proportional hazards model, with a strata statement accounting for matched pairs, estimated the relative hazard of death within 1 year. The proportional hazards assumption for referral status was assessed by creating an interaction term with time.

RESULTS

There were 265 patients who met the inclusion criteria for the primary analysis (Table 1). Overall, patients were on average 77 years old (standard deviation [SD]: 5.26 years), predominantly non-Hispanic White (89.43%), and married or in a domestic partnership (65.66%). Patients primarily were diagnosed with lung cancer (26.89%) or gastrointestinal cancer (24.24%). Nearly one-third of patients were receiving a monoclonal antibody treatment (31.8%), and just over two-thirds were receiving chemotherapy (68.44%).

TABLE 1.

Demographic characteristics of older individuals with advanced cancer with a physical function or functional status impairment on the geriatric assessment (n = 265).

		Total sample		Not referred to rehabilitation (n = 182)		Referred to rehabilitation (n = 83)
	Category	N	%	N	%	N	%	p-value
Demographic variable
Age	Mean (SD)	76.8 (5.3)		76.2 (4.7)		78.1 (6.1)
	Range	70–96		70–90		70–96
Age categorically	70–79 years	199	75.1	145	79.7	54	65.1	0.01
	80 years or older	66	24.9	37	20.3	29	34.9
Race/ethnicity	Non-Hispanic White	237	89.4	164	90.1	73	87.9	0.6
	Other	28	10.6	18	9.9	10	12.1
Gendera	Male	138	52.1	101	55.5	37	44.6	0.1
	Female	127	47.9	81	44.5	46	55.4
Education	Less than high school or high school graduate	138	52.1	94	51.7	44	53.0	0.8
	Some college or above	127	47.9	88	48.3	39	47.0
Income	Less than or equal to $50K	138	52.1	94	51.7	44	53.0	0.2
	More than $50K	80	30.2	51	28.0	29	34.9
	Declined to answer	47	17.7	37	20.3	10	12.1
Marital	Married or domestic partnership	174	65.7	125	68.7	49	59.0	0.1
	Single, separated, widowed, divorced	91	34.4	57	31.3	34	41.0
Living arrangement	Independent living (one or more story)	118	44.5	83	45.6	35	42.2	0.8
	Independent living (one story)	137	51.7	93	51.1	44	53.0
	Others	10	3.8	6	3.3	4	4.8
Living situation	Lives alone	52	19.6	32	17.6	20	24.1	0.2
	Lives with others	213	80.4	150	82.4	63	75.9
Health-related quality of life (FACT-G)	Mean (SD)	80.1 (14.9)		82.01 (14.0)		75.89 (16.2)		<0.01
Geriatric assessment domains
Comorbiditiesb	<3 comorbidities and none with a great deal of interference	88	33.3	59	32.6	29	34.9	0.7
	≥3 comorbidities or any with a great deal of interference	176	66.7	122	67.4	54	65.1
MOS social supportb	Not impaired	195	73.9	140	77.4	55	66.3	0.06
	Impaired (at least one item with 1, 2, or 3 value)	69	26.1	41	22.6	28	33.7
Generalized Anxiety	Not impaired	245	92.5	173	95.1	72	86.7	0.02
Disorder-7	Impaired (total score ≥10)	20	7.5	9	4.9	11	13.3
Geriatric Depression Scale	Not impaired	205	77.4	145	79.7	60	72.3	0.2
short form	Impaired (total score ≥5)	60	22.6	37	20.3	23	27.7
Blessed orientation	Not impaired	257	97.0	179	98.4	78	94.0	0.05
memory concentration	Impaired (total score ≥11)	8	3.0	3	1.6	5	6.0
Mini-Cogb	Not impaired	170	64.4	129	71.3	41	49.4	<0.01
	Impaired (recalled 0 words or 1–2 words with abnormal clock)	94	35.6	52	28.7	42	50.6
Mini nutrition assessment	Not impaired	104	39.3	79	43.4	25	30.1	0.04
	Impaired (total score ≤11)	161	60.7	103	56.6	58	69.9
Body mass index	Not impaired	236	89.1	166	91.2	70	84.3	0.1
	Impaired (BMI < 21 kg/m2)	29	10.9	16	8.8	13	15.7
Use of high-risk drugsb	Patient does not take any high-risk drug	118	44.7	74	40.9	44	53.0	0.07
	Patient takes at least one high-risk drug	146	55.3	107	59.1	39	47.0
Polypharmacy	<5 prescription medications taken regularly	106	40.0	63	34.6	43	51.8	0.01
	≥5 or more prescription medications taken regularly	159	60.0	119	65.4	40	48.2
Short physical performance	Not impaired (total score ≥10)	53	20.0	43	23.6	10	12.1	0.03
battery	Impaired (total score ≤9)	212	80.0	139	76.4	73	87.9
Fall history	No fall within past 6 months	184	69.4	131	72.0	53	63.9	0.2
	One or more falls within 6 months	81	30.6	51	28.0	30	36.1
Time up and go	Not impaired (≤13.5 s)	171	64.5	131	72.0	40	48.2	<0.01
	Impaired (>13.5 s or did not attempt/failed)	94	35.5	51	28.0	43	51.8
Activities of daily living	Not impaired (six items without dependency)	192	72.5	142	78.0	50	60.2	<0.01
	Impaired (at least one item with dependency)	73	27.5	40	22.0	33	39.8
Instrumental activities of daily living	Not impaired (seven items without dependency	102	38.5	72	39.6	30	36.1	0.6
	1 = Impaired (at least one item with dependency)	163	61.5	110	60.4	53	63.9
NCCN distress	Not impaired (<4)	172	65.7	124	68.5	48	59.3	0.15
thermometerd	Impaired (≥4)	90	34.3	57	31.5	33	40.7
Clinical information
Cancer typea	Gastrointestinal	64	24.2	43	23.8	21	25.3	0.9
	Lung	71	26.9	50	27.6	21	25.3
	Other	129	48.9	88	48.6	41	49.4
Currently receiving	No	83	31.6	53	29.3	30	36.6	0.2
chemotherapyc	Yes	180	68.4	128	70.7	52	63.4
Currently receiving	No	244	93.86	169	93.9	75	93.8	0.9
radiation therapye	Yes	16	6.15	11	6.1	5	6.2
Currently receiving	No	180	68.2	131	72.4	49	59.0	0.03
monoclonal antibodiesb	Yes	84	31.8	50	27.6	34	41.0
	No	225	85.2	154	85.1	71	85.5	0.9
Currently receiving	Yes	39	14.8	27	14.9	12	14.5
hormonal therapyb
Karnofsky performance	Impaired	62	23.5	30	16.5	32	39.0	<0.01
status—clinician-ratedb	Normal	202	76.5	152	83.5	50	61.0

Note: Bolded values denote a statistically significant (p < 0.05) outcome.

Abbreviations: BMI, body mass index; FACT-G, Functional Assessment of Cancer Treatment—General; MOS, Medical Outcomes Study; NCCN, National Comprehensive Cancer Network.

^aRepresents self-reported gender (social construct) as selected by the participant.

^bSample size = 264.

^cSample size = 263.

^dSample size = 262.

^eSample size = 260.

Predictors of oncologist-initiated discussion about or implemented referral to rehabilitation services

There were significant differences (p < 0.05) in the characteristics between the group of patients who received an oncologist-initiated discussion about or referral to rehabilitation compared with those who did not (Table 1). Patients who received an oncologist-initiated discussion about or referral to rehabilitation had a significantly higher proportion of individuals 80 years or older (34.94% vs. 20.33%) and currently receiving monoclonal antibodies (40.96% vs. 27.62%). Of the GA domains, patients who received an oncologist-initiated discussion about or implemented referral had a significantly higher proportion of individuals with anxiety (13.25% vs. 4.95%), impaired cognition (50.60% vs. 28.73%), and impaired nutrition (69.88% vs. 56.59%). Patients without discussion or referral had a higher proportion with polypharmacy compared with those who did. Univariable analysis of demographic, clinical, and GA domain-related characteristics associated with rehabilitation referral can be found in Supplementary Table S3.

Multivariable modeling revealed several factors that were significantly associated with an oncologist-initiated discussion about or referral to rehabilitation (Table 2). Specifically, patients with impaired cognition had higher odds of an oncologist-initiated discussion about or implemented referral to rehabilitation compared with those with normal cognition (OR 2.25; 95% [CI]: 1.19–4.26). Compared with patients without polypharmacy, patients with polypharmacy had lower odds of receiving an oncologist-initiated discussion about or referral to rehabilitation (OR 0.31; 95% CI: 0.16–0.59). Patients with a clinician-rated Karnofsky Performance Status of any disability level had 2.86 higher odds (95% CI: 1.37–5.97) of receiving an oncologist-initiated discussion about or referral to rehabilitation compared with individuals with normal performance status. Lastly, patients who received monoclonal antibody use had higher odds of an oncologist-initiated discussion about or referral to rehabilitation (OR 1.95; 95% CI: 1.02–3.71) compared with those who did not. Sensitivity analyses revealed that the interaction terms between treatment type and cancer type variables were not significant. Model fit or previously identified associations did not change significantly when adding the interaction term.

TABLE 2.

Adjusted odds ratios and 95% confidence intervals for an oncologist-initiated discussion about or referral to rehabilitation after a functional status or physical performance impairment on the geriatric assessment (n = 262).

			95% confidence interval
Variable	Category	O6R	Lower limit	Upper limit	p-value
Age	70–79	-	-	-	-
	80+	1.62	0.83	3.17	0.16
Gendera	Male	-	-	-	-
	Female	1.14	0.61	2.10	0.69
Comorbidity status	<3 comorbidities and none with a great deal of interference	-	-	-	-
	≥3 comorbidities or any with a great deal of interference	0.67	0.35	1.30	0.24
MOS social support	Not impaired	-	-	-	-
	Impaired (at least one item with 1, 2, or 3 value)	1.75	0.89	3.46	0.11
GAD-7	Not impaired	-	-	-	-
	Impaired (total score ≥10)	2.00	0.69	5.82	0.20
Mini-Cog	Not Impaired	-	-	-	-
	Impaired (recalled 0 words or 1–2 words with abnormal clock)	2.25	1.19	4.26	0.01
MNA	Not impaired	-	-	-	-
	Impaired (total score ≤11)	1.28	0.67	2.44	0.46
Polypharmacy	Less than five prescription medications taken regularly	-	-	-	-
	Five or more prescription medications taken regularly	0.31	0.16	0.59	<0.01
Short physical performance	Not impaired (total score ≥10)	-	-	-	-
battery	Impaired (total score ≤9)	1.42	0.58	3.49	0.45
Time up and go (TUG)	Not impaired (≤13.5 s)	-	-	-	-
	Impaired (>13.5 s or did not attempt/failed)	1.44	0.72	2.87	0.30
Activities of daily living	Not impaired (six items without dependency)	-	-	-	-
	Impaired (at least one item with dependency)	1.71	0.82	3.55	0.15
Karnofsky performance status—clinician-rated	Disabled (unable normal, occasional assistance, considerable assistance, disabled)	2.86	1.37	5.97	0.01
	Normal, minor symptoms, normal with effort	-	-	-	-
Cancer type	Gastrointestinal	-	-	-	-
	Lung	0.79	0.34	1.85	0.57
	Other	0.93	0.44	1.99	0.87
Monoclonal antibody therapy	No	-	-	-	-
	Yes	1.95	1.02	3.71	0.04

Abbreviations: GAD-7, Generalized Anxiety Disorder-7; MNA, Mini Nutrition Assessment.

^aRepresents self-reported gender (social construct) as selected by the participant.

Association between implemented rehabilitation referral only with function and HRQoL

Of the 265 patients with data on rehabilitation referral, 250 had implementation status recorded (68 had an implemented referral, 182 did not have an implemented referral). Sensitivity analyses revealed that patients who received an oncologist-initiated discussion about rehabilitation but did not have an implemented referral had significantly lower income, a lower proportion receiving chemotherapy, greater declines in ADL, IADL, and HRQoL, and differed in their cancer types compared with individuals who received an implemented referral (Supplementary Table S4). Propensity score matches were identified for 53 of the referred patients. Of these, 13 referred and 4 non-referred patients were missing data on outcomes at 3 months, leading to a final sample of 36 matched pairs (n = 72) (Supplementary Tables S5 and S6; Supplemental Figure S2a,b).

As shown in Table 3, those with implemented referral had lower odds of decline in ADL and IADL but greater odds of reporting a decline in HRQoL compared with non-referred patients. Estimates from the same model in the unmatched sample show diminished protective effect for decline in function and greater odds of decline in HRQoL (Supplementary Table S7). The results from propensity score analysis demonstrated a nonsignificant shift toward an attenuated effect of implemented rehabilitation referral on functional decline (ADL: before matching OR = 0.77, 95% CI: 0.36–1.63; after matching OR = 0.30, 95% CI 0.08–1.09. IADL: before matching OR = 1.16, 95% CI: 0.52–1.85; after matching OR = 0.64, 95% CI 0.29–2.09).

TABLE 3.

Odds ratios and 95% confidence intervals from conditional logistic regression for functional and quality of life outcomes for propensity score matched patients who have received an implemented referral compared with patients without an oncologist-initiated discussion about or implemented referral to rehabilitation (n = 72).

		95% confidence interval
Outcome	OR	Lower limit	Upper limit	p-value
Decline in activities of daily living (ADL)	0.30	0.08	1.09	0.07
Decline in instrumental ADL (14 point scale)	0.64	0.25	1.64	0.35
Decline in FACT-Ga	1.20	0.52	2.78	0.67

Abbreviation: FACT-G, Functional Assessment of Cancer Therapy-General.

^aN = 70 due to one additional patient missing FACT-G.

Association between implemented rehabilitation referral only and survival

All 53 matched pairs had data for one-year overall survival. Compared with non-referred patients, those with an implemented referral had 1.18 times the relative hazard of mortality; however, this was not statistically significant (95% CI 0.62–2.25; p = 0.62) (Figure 2). Comparing this result to estimates from the unmatched sample, there was a greater relative hazard of mortality associated with referral prior to propensity score matching (HR 1.44, 95% CI 0.93–2.22; p = 0.10, Supplemental Figure S3).

FIGURE 2.

One-year overall survival by referral status among propensity score matched patients who have received an implemented referred referral compared with patients without an oncologist-initiated discussion about or implemented referral to rehabilitation (n = 106).

DISCUSSION

The GA offers a natural triage pathway to refer older adults with advanced cancer who experience physical or functional deficits to rehabilitation services. From the GA, approximately 93% of older adults with advanced cancer in our sample demonstrated physical and/or functional deficits, yet only 25% received an implemented rehabilitation referral. Main findings indicate that an oncologist-initiated discussion about and/or implemented referral to rehabilitation services were more likely among those with cognitive impairment, clinician-reported disability, and those treated with monoclonal antibodies. In contrast, odds of referral to rehabilitation were significantly lower among individuals with polypharmacy. Although not statistically significant, patients who received implemented referral had 60% lower odds of decline in performance of ADL in the 3 months after GA than those who did not discuss or have an implemented referral to rehabilitation. Altogether, these findings indicate key considerations for access and referral to rehabilitation, signal potential benefits of rehabilitation services, and underscore the need for future research related to rehabilitation access and delivery in this population.

Understanding patterns in GA-driven referral to rehabilitation may clarify current care delivery practices related to addressing physical performance and functional deficits with rehabilitation services. In the adjusted model, the odds of an oncologist-initiated discussion about or referral to rehabilitation were higher among those with clinician-reported disability, cognitive impairment, and exposure to monoclonal antibodies. Whereas disability and cognitive impairment are two of the consistent indications for referral to rehabilitation services,^37,38 this study identified a relationship between monoclonal antibody use and referral. One reason for this outcome may be due to lower rates of grade ≥3 toxicity with monoclonal antibodies among older patients with cancer with GA-related impairments.³⁹ Thus, oncologists may judge that older adults receiving this treatment are more likely to tolerate rehabilitation services and be more likely to initiate a referral. Additional prospective research is warranted to clarify this association and emerging relationship.

The results may also help reduce undertreatment of physical or functional impairments by identifying patients who are less likely to be referred. For instance, individuals with polypharmacy were nearly 70% less likely to receive a referral to rehabilitation. Polypharmacy is often associated with comorbidity, frailty, and poorer physical function among older adults with cancer.⁴⁰ Thus, providers may consider deprescribing options as a first-line intervention to reduce polypharmacy-related functional decline (low blood pressure, peripheral neuropathy, etc.)⁴⁰ Alternatively, providers may refrain from referring these patients with polypharmacy because they judge their prognosis is less than 12 months or their disease is unlikely to be cured. Consistent with the current findings, Spill et al. found that medical oncologists seldom refer individuals with advanced cancer to rehabilitation if prognosis is less than 12 months.⁴¹ However, Gorzelitz et al. found that older adults with breast cancer with greater numbers of comorbidity and medical complexity were more likely to utilize rehabilitation services compared with those with lower medical complexity.²⁸ Despite mixed evidence, rehabilitation interventions can be tailored to support patient safety and medical complexity in this population.⁴² Additional research clarifying this relationship is warranted.

In advanced cancer, older adults’ priorities often shift from curative toward those that prioritize functional status and HRQoL.⁴³ Although changes in function are often inevitable due to cancer and aging, slowing such declines may enhance HRQoL. There were no statistical differences in functional decline at 3 months between patients in our sample who received a referral and those who did not. This may be driven by multiple factors. First, among patients with impaired physical and/or functional status deficits, oncologist judgment, over GA-driven recommendations alone, may have led to greater likelihood of referral for frail patients who were more likely to decline during follow-up. This is supported by the attenuated effect estimates after propensity score matching and increased odds for referral based on clinician-reported KPS score. Second, data on subsequent utilization of rehabilitation were not available; thus, these findings do not consider actual use of or dose of rehabilitation received. Rehabilitation dose has been previously documented as an important factor for improving functional outcomes in other patient populations.⁴⁴ Recent investigations into the relationship between outpatient rehabilitation use and outcomes in this population indicate significant improvement in functional outcomes.^45,46 Lastly, although the measures of ADL and IADL in this study are widely used and validated in this population, they focus on need for assistance and may not reflect subtle changes that are meaningful to patients. For example, the ADL score would not change for a patient who improves from requiring complete assistance to set up assistance for dressing, although such progress denotes improved independence. The analyses were designed to estimate preliminary effect sizes that will guide future research. As such, future research should be designed to capture dose, and content of rehabilitation utilization after referral and more sensitive measures of function.

Finally, it is important to note that only 30% of older adults with advanced cancer who had a GA-driven referral recommendation reported an oncologist-initiated discussion about rehabilitation, with only 25% receiving an implemented referral. Although these findings result from data from 2014 to 2017, these proportions are similar,^7,28 and in some cases higher,² to prevalence of rehabilitation uptake among older adults with cancer over the past 10 years. Furthermore, the referrals may better align with availability of resources in community oncology settings. Understanding patient and provider decision-making after review of GA-referral recommendations is warranted. In the present study, oncologists had autonomy for if and how they sought to use the GA with their patients. Thus, oncologist’s may have used personal experience, perception of function, and rehabilitative outcomes (e.g., KPS score) or other clinical decision-making processes over use of the GA recommendations. For example, risk perceptions of specific impairments might influence providers to refer only some older adults with advanced cancer. Understanding the nuance underlying personal judgment will help ensure that equitable access to rehabilitation is made when patient need (e.g., physical or functional impairment) dictates care delivery (e.g., rehabilitation services).⁴⁷ Qualitative studies to evaluate differences between those who may only discuss rehabilitation opportunities versus those who receive an implemented referral are also warranted. Future work should investigate provider- and organizational-level factors that may influence decisions to pursue rehabilitation referral including availability of rehabilitation services, accessibility or services, and provider perspective of rehabilitation.

Limitations

This study represents one of the first opportunities to study GA-driven referral patterns to rehabilitation in community oncology settings. Furthermore, it captures supportive care referral patterns among older adults with advanced cancer who have been underrepresented in clinical trials⁴⁸ and rehabilitation research.⁴⁹ However, these findings should be reviewed in light of limitations. The small sample size, inclusive of cases with missing data, may limit generalizability and reliability of findings beyond our sample. Specifically, because the overlap in propensity score values between the two groups is necessary for finding propensity score match and there were likely intrinsic differences in characteristics of patients who received or did not receive referral, the propensity score match was found only for a smaller sub-cohort of patients. Current data availability did not capture potential rehabilitation discussions or referrals made not in reference to the GA-driven recommendations. Thus, referrals could be under-reported. Our findings do not consider alternate interventions that may address function, such as exercise oncology programming, or the specific setting (e.g., outpatient, home health) in which rehabilitation referral was made. Future research should evaluate access and delivery of exercise oncology programming in this population as well as frequency of rehabilitation setting. Finally, our analysis does not reflect outcomes based on services utilized. Data regarding referral were documented as a discussion and/or implemented referral. Thus, we cannot firmly conclude that individuals formally utilized rehabilitation services. Our analyses were designed to estimate preliminary effect sizes that will guide future research to capture patterns in rehabilitation utilization after referral, as well as characterize changes in clinical outcomes associated GA-driven rehabilitation recommendations.

CONCLUSION

This study found that older adults with advanced cancer are more likely to receive an oncologist-initiated discussion about or referral to rehabilitation after detection of cognitive impairment, clinician-reported disability, and exposure to monoclonal antibodies. Compared with non-referred patients, patients who received an implemented referral to rehabilitation did not statistically significantly differ in change in ADL, IADL, or HRQoL over time. Further research is needed on underlying decision-making, risk perception, and barriers to referral and utilization rehabilitation services among older adults with advanced cancer.

Key points

• Among older adults with advanced cancer who experience functional limitations, cancer treatment, polypharmacy, cognition, and disability status likely influence oncologists’ decision to refer to rehabilitation services.

• Understanding patterns in geriatric assessment-driven referral to rehabilitation services may clarify current care delivery practices related to addressing physical performance and functional deficits among older adults with advanced cancer.

Why does this paper matter?

Older adults with advanced cancer experience functional disability that warrants rehabilitation services; however, evidence indicates inconsistencies in referral. The geriatric assessment offers a natural triage pathway to successfully refer older adults with advanced cancer to rehabilitation services. Understanding patterns in geriatric assessment-driven referral to rehabilitation services may clarify current care delivery practices related to addressing physical performance and functional deficits among older adults with advanced cancer. Among older adults with advanced cancer who experience functional limitations, cancer treatment, polypharmacy, cognition, and disability status likely influence oncologists’ decision to refer to rehabilitation services. Altogether, these findings indicate key considerations for access and referral to rehabilitation in this population, signal potential benefits of rehabilitation services for this population, and underscore the need for future research related to rehabilitation access and delivery.