What Drove Clients’ Decisions to Pause Personal Homecare Services Before and During the Pandemic?

Abstract

Background:

With an increasing reliance on homecare and a scarcity of providers, there is potential to gain insight from existing administrative data to optimize planning and care delivery. To enable more accurate predictions of service use, it is important to understand the degree to which various factors influence clients’ difficult decisions to temporarily pause their receipt of necessary homecare services.

Objectives:

We utilized a large, longitudinal, administrative dataset to examine the relative effects of client-level factors on the outcomes of (1) placing a hold on homecare services and (2) the length of a homecare service hold, through stratified regression analyses separated by pre-, early-, and mid-pandemic periods.

Design:

Descriptive summaries of the samples consisted of graphical representation and frequencies (proportions) or means. The relationship between client sociodemographic and homecare utilization factors on the service hold initiation and length were evaluated using mixed-effects logistic and linear regression, respectively, stratified by pre-, early-, and mid-pandemic periods. Odds ratios (OR) for hold initiation and exponentiated estimates for hold length were calculated with corresponding 95% confidence intervals.

Results:

Findings provide a better understanding of the decisions made by a large sample of homecare clients to pause their homecare services in pre-, early-, and mid-pandemic scenarios. Frequency and length of service holds more than doubled in the early-pandemic period; although hold frequencies then returned to pre-pandemic rates, hold durations remained slightly longer. There were notable differences over time, but generally, clients with higher care needs had a reduced likelihood of placing a hold on homecare services. Shorter homecare tenure and previously cancelling individual homecare visits were also good indicators of future service decisions.

Conclusion:

Findings are relevant for organizations providing homecare services, policymakers, and those interested in predicting homecare utilization for resource allocation planning with the goal of optimal care delivery.

Keywords

health service utilization homecare healthcare delivery COVID-19 mixed effects regression analysis client/patient choice personal support work healthcare planning

Introduction

Global population ageing has contributed to higher demand for homecare services over time, with forecasts projecting continued growth into the future.¹ However, despite the substantial need for homecare services as an essential component of healthcare systems, shocks like the COVID-19 pandemic can disrupt expected homecare utilization,^2,3 leading to challenges for care delivery planning to ensure that individuals are receiving the care that they need. During the COVID-19 pandemic, the initial drop in homecare services^2-5 was primarily driven by a decrease in demand rather than a reduced labor supply,^5,6 and was most pronounced for recipients of personal support services.² Reductions in the receipt of home care services during the pandemic have been associated with negative impacts for both the individuals seeking care and for family and friends who have stepped in as caregivers to provide necessary care.^5,7-9 Changes in demand can also negatively impact homecare providers, as client caseload instability affects their work and income.¹⁰

It is common for homecare clients receiving personal support services to choose to temporarily pause care,⁶ yet the reasons why clients make these choices are not well understood. Existing literature on contributing factors for clients’ decisions to reduce homecare services is relatively limited, largely qualitative, and is often specific to sub-populations such as individuals with dementia.^2,9,11,12 Previous qualitative literature has highlighted concerns about infection risk, increased support from family members and friends, and changes in homecare worker availability as influencing decisions to pause receipt of formal care services.^2,9,11,12 Client health is expected to play a role in homecare utilization, with one quantitative study showing that homecare recipient’s health (across a variety of factors) tended to be worse at the beginning of the COVID-19 pandemic vs pre-pandemic.²

Given the limited and largely descriptive nature of existing studies which relate to homecare service holds, additional analysis is required to understand how various client and service characteristics contribute to clients’ decisions to place and maintain homecare service holds. With the growth in demand for homecare forecasted to continue, gaining a better understanding of homecare service patterns during pandemic and non-pandemic periods is increasingly important to allow for informed decision-making around resourcing in this sector. The objectives of this research were to determine the relative significance of socio-demographic and care utilization factors in affecting a homecare client’s decision to place their homecare personal support services on hold, and the durations of homecare service holds.

Methods

Data

This retrospective open cohort study leveraged longitudinal administrative quantitative data from a large homecare organization operating in the Greater Toronto Area of Ontario, Canada. The sample included all adult clients (aged 18 and over) who were receiving or had placed a hold on receipt of publicly funded personal support worker (PSW) services (excluding palliative care) from this organization between January 1, 2019 and June 30, 2021. PSWs provide the majority of paid homecare services, often providing care for activities of daily living (ADLs), such as helping clients to eat and bathe, as well as instrumental activities of daily living (IADLs), such as meal preparation and laundry. The open cohort design is reflective of the nature of the home care recipient population and allowed for clients to enter or exit the sample in any given week as they began or ended homecare services with the organization. The data were stratified by pre-pandemic (January 1, 2019-February 25, 2020), early-pandemic (February 26, 2020-August 31, 2020), and mid-pandemic (September 1, 2020-June 30, 2021) periods, characterized by local COVID-19 waves¹³ for all analyses.

Measures

Dependent Variables

Two outcome variables were of interest: (1) service hold initiation and (2) length of service hold. A client-initiated service hold was defined as a temporary pause of PSW services, initiated by the client or caregiver and unrelated to a hospital admission, that lasted two or more days. The length of service hold was defined as the number of days between hold initiation and resumption of services. The objectives of this paper were to characterize voluntary client-initiated holds, therefore, holds due to hospitalization were excluded as these are generally out of the client’s control.

Independent Variables

The dataset included client-level socio-demographic and homecare utilization factors. Socio-demographic factors included age, sex, primary spoken language (closest available proxy for ethnic/cultural background), marital status, number of contacts on file with the care provider organization (closest available proxy for depth of social support), acuity of care needs, and region. Homecare utilization factors included congregate setting, number of care programs, type of care, service intensity, client tenure with organization, whether the client had previously chosen to place a service hold (for reasons other than hospitalization), and whether, in the most recent week care was received, there had been client-initiated cancellation of individual service visits, missed care events wherein a scheduled visit was not fulfilled by the provider organization, or the introduction of a new PSW from whom the client had never previously received care. The variable for number of contacts on file with the care provider organization captures the number of individuals who were authorized to be contacted if needed regarding a given client’s care (this often includes a client’s spouse, child, friend, or other individual). Geographic area was used to control for differences in policy and administrative practices between three distinct health regions, as defined by the Home and Community Care Support Services (HCCSS) – the quasi-governmental agencies responsible for coordinating homecare services in Ontario during the study period. Receiving care in a congregate setting (eg, apartment building with multiple homecare clients, retirement home, long-term care facility) was also included in the model. Acuity of client care needs was categorized as “low,” “medium,” or “high” based on “emergency response level,” which is a variable received as part of the homecare service referral package reflecting the urgency of a client’s care needs. Client care needs were further captured by 1) whether the client received care for IADL tasks (eg, meal preparation, laundry, etc.) in addition to ADL tasks (eg, bathing, toileting, etc.), and 2) whether the client received care for multiple program types (ie, providing personal care, helping with homemaking, or offering caregiving relief through respite care). The average number of visits that a client received over the 4 weeks prior to a hold initiation was used to characterize service intensity.

Analysis Strategy

Considering the impact of the pandemic and the variation in health policies during the study period, models were stratified by three periods (pre-, early-, and mid-pandemic). Descriptive and inferential statistical methods were applied to the data within each separated period. Service hold initiation was summarized and analyzed at the individual and weekly longitudinal level. Descriptive summaries of the samples consisted of graphical representation and frequencies (proportions) or means. Generalized Estimating Equations (GEE; P < .05) were used to assess the significance of predictor variables on the outcome variable for service hold initiation (binomial variable: “Initiated a service hold” vs “Did not initiate a service hold”). GEE were also used to assess the differences in proportions or means across all periods (pre- vs early-pandemic, pre- vs mid-pandemic, and early- vs mid-pandemic). GEE were utilized to account for non-independence between periods,^14,15 recognizing that an individual could contribute data during multiple weeks and across multiple stratified periods. The relationship between client sociodemographic and homecare utilization factors on the service hold initiation and length were evaluated using mixed-effects logistic and linear regression, respectively. The outcome variable of hold length was log transformed due to the skewed nature of the data and results were presented as exponentiated estimates for ease of interpretation. Odds ratios (OR) for hold initiation and exponentiated estimates for hold length were calculated with corresponding 95% confidence intervals. We also compared the regression coefficients between the three stratified models, following from Clogg et al,¹⁶ to assess the impact of each predictor variable on the outcome across the time periods, and significant findings were reported in-text. All analyses were performed in RStudio version 4.2.2. All tests of significance used an a priori threshold of P < .05. Sensitivity analyses were conducted to test the robustness of the model, including iterative removal of each predictor variable which revealed consistent direction and strength of significance were maintained even with the removal of one of the predictors. Variance inflation factors (VIFs) were obtained to measure multicollinearity with an a priori maximum cut-off of 10;¹⁷ the highest VIF value is reported in table notes.

Approval for this study was provided by the University of Toronto Research Ethics Board (REB# 40086). The need for consent was waived by the ethics committee for the use of this secondary data due to satisfying all conditions outlined in Article 5.5A of the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans.¹⁸

Results

Sample Characteristics for Each Period

Overall, 16 780 unique clients received home care services during the study period (January 1, 2019-June 30, 2021). Demographic and service utilization characteristics of the cohort are summarized in Table 1 for each of the three stratified time periods. Comparisons across periods revealed significant differences in select socio-demographic and care utilization variables over time. Specifically, clients receiving care during the pandemic received more visits, had longer average tenures, and had higher number of contacts on file (early-pandemic vs pre-pandemic), compared to the pre-pandemic home care population.

Table 1.

Characteristics of client weeks for each period.

	Pre-Pandemic (60 weeks); N = 387 992	Early-Pandemic (27 weeks); N = 147 577	Mid-Pandemic (44 weeks); N = 252 654	Pre- vs early-²	Pre- vs mid-²	Early- vs mid-²
Client characteristics	Mean (SD); n (%)	Mean (SD); n (%)	Mean (SD); n (%)	P-value	P-value	P-value
Demographics
Age (years)	78.3 (16.2)	78.5 (15.8)	78.5 (15.7)	<.001	<.001	<.001
Age (years)
Under 60 (ref)	45 644 (11.8%)	16 596 (11.2%)	28 069 (11.1%)
60-69	39 162 (10.1%)	15 461 (10.5%)	27 267 (10.8%)	.020	.001	.062
70-79	73 800 (19.0%)	28 702 (19.4%)	50 394 (19.9%)	.002	<.001	.180
80-89	144 519 (37.2%)	54 445 (36.9%)	90 362 (35.8%)	<.001	<.001	.390
90 and above	84 867 (21.9%)	32 373 (21.9%)	56 562 (22.4%)	<.001	<.001	<.001
Sex
Female (ref)	257 736 (66.4%)	97 122 (65.8%)	167 609 (66.3%)
Male	130 256 (33.6%)	50 455 (34.2%)	85 045 (33.7%)	.110	.340	.529
Languages
English (ref)	301 647 (77.7%)	116 827 (79.2%)	201 503 (79.8%)
Other	86 345 (22.3%)	30 750 (20.8%)	51 151 (20.2%)	<.001	<.001	.001
Marital status
Single (ref)	199 223 (51.3%)	73 923 (50.1%)	126 149 (49.9%)
Partnered	98 529 (25.4%)	35 481 (24.0%)	59 048 (23.4%)	.162	.003	.044
Unknown	90 240 (23.3%)	38 173 (25.9%)	67 457 (26.7%)	<.001	<.001	.299
Number of contacts on file
0-1 (ref)	96 899 (25.0%)	34 573 (23.4%)	66 941 (26.5%)
2	132 282 (34.1%)	49 771 (33.7%)	86 233 (34.1%)	.065	.003	<.001
3 or more	158 811 (40.9%)	63 233 (42.8%)	99 480 (39.4%)	<.001	<.001	<.001
Acuity of care needs
Low risk (ref)	141 041 (36.4%)	52 070 (35.3%)	94 317 (37.3%)
Medium risk	120 568 (31.1%)	47 380 (32.1%)	88 051 (34.9%)	<.001	<.001	.027
High risk	96 160 (24.8%)	39 865 (27.0%)	65 327 (25.9%)	<.001	.333	.001
Unknown	30 223 (7.79%)	8262 (5.6%)	4959 (1.9%)	<.001	<.001	<.001
Health region
Toronto Central (ref)	157 859 (40.7%)	61 442 (41.6%)	103 830 (41.1%)
Central East	114 168 (29.4%)	42 652 (28.9%)	74 479 (29.5%)	.360	.504	.169
Central	115 965 (29.9%)	43 483 (29.5%)	74 345 (29.4%)	.290	.540	.675
Home care utilization factors
Care setting
Non-congregate (ref)	313 523 (80.8%)	116 593 (79.0%)	192 318 (76.1%)
Congregate	74 469 (19.2%)	30 984 (21.0%)	60 336 (23.9%)	.001	<.001	<.001
Number of programs
1 (ref)	359 288 (92.6%)	138 209 (93.7%)	239 025 (94.6%)
2 or more	28 704 (7.4%)	9368 (6.3%)	13 629 (5.39%)	<.001	<.001	<.001
Care type
ADL (ref)	145 483 (37.5%)	54 141 (36.7%)	98 298 (38.9%)
ADL + IADL	242 509 (62.5%)	93 436 (63.3%)	154 356 (61.1%)	.329	.002	<.001
Service intensity	4.7 (4.6)	4.9 (4.8)	5.0 (4.8)	<.001	<.001	<.001
Tenure (weeks)	124.8 (162.5)	134.2 (167.7)	137.2 (168.3)	<.001	<.001	<.001
Tenure
Under 3 months (ref)	82 863 (21.4%)	28 136 (19.1%)	40 906 (16.2%)
3-6 months	32 655 (8.4%)	11 583 (7.8%)	23 422 (9.27%)	<.001	.506	<.001
6-12 months	52 608 (13.6%)	18 086 (12.3%)	34 095 (13.5%)	<.001	<.001	<.001
12-24 months	72 180 (18.6%)	28 093 (19.0%)	45 443 (18.0%)	<.001	<.001	<.001
Over 24 months	147 686 (38.1%)	61 679 (41.8%)	108 788 (43.1%)	<.001	<.001	<.001
Placed a previous hold
No (ref)	353 345 (91.1%)	132 914 (90.1%)	227 575 (90.1%)
Yes	34 647 (8.93%)	14 663 (9.9%)	25 079 (9.9%)	<.001	<.001	<.001
Visit from a new PSW ¹
No (ref)	307 999 (79.4%)	123 459 (83.7%)	214 206 (84.8%)
Yes	79 993 (20.6%)	24 118 (16.3%)	38 448 (15.2%)	<.001	<.001	<.001
Cancellations ¹	0.140 (0.479)	0.145 (0.531)	0.143 (0.499)	<.001	<.001	.689
Missed cares ¹	0.013 (0.129)	0.006 (0.092)	0.005 (0.093)	<.001	<.001	.007

Reference categories for comparison are labeled as (ref).

Factors reflect the most recent week that care was received (eg, if on a hold, the last week that care was received included a cancellation).

Comparison across pre-, early, and mid-pandemic time periods assessed through Generalized Estimating Equations.

Of the clients who received PSW services, 5228 (31.2%) initiated a service hold during the study period. Of these 5228 clients, 80% (n = 4187) initiated a single hold, while 20% (n = 1041) initiated multiple service holds (range: 2-9 holds). Most clients who initiated a hold were female (63.4%) with an average age of 76.7 years. As shown in Figure 1, hold initiation and duration more than doubled in the early-pandemic period, before decreasing back towards pre-pandemic rates in the mid-pandemic period. Early-pandemic, there was a 117% increase in the number of average weekly service holds initiated, compared to pre-pandemic (P < .05). On average, these holds lasted 26 days longer than pre-pandemic holds (P < .001). By mid-pandemic, trends in hold initiation returned near pre-pandemic rates, but these holds tended to last longer (averaging 5 days longer than pre-pandemic, P < .001).

Figure 1.

Frequency of new client-initiated holds over time.

Demographic and Homecare Utilization Factor Effects on the Outcomes of Hold Initiation and Hold Length

Although Figure 1 indicates a general return to pre-pandemic hold rates, clients’ hold-related decision making had not returned to the pre-pandemic baseline by mid-pandemic. There were several significant differences over time in how socio-demographic and care-related factors correlated with the decision to place homecare services on hold, most of which were realized at the onset of the pandemic (Table 2).

Table 2.

Characteristics of total client weeks with and without a homecare service hold, stratified by period.

	Pre-pandemic (60 weeks)		Early-pandemic (27 weeks)		Mid-pandemic (44 weeks)
	Weeks with and without a non-hospital service hold
	No hold	Hold	No hold	Hold	No hold	Hold
	N = 385 495	N = 2497	N = 145 148	N = 2429	N = 250 947	N = 1707
Client characteristics	Mean (SD); n (%)		Mean (SD); n (%)		Mean (SD); n (%)
Demographics
Age (years)	78.4 (16.2)	76.0 (16.4)***	78.6 (15.8)	77.3 (16.7)**	78.5 (15.7)	77.0 (15.6)***
Age (years)
Under 60 (ref)	45 306 (11.8%)	338 (13.5%)	16 289 (11.2%)	307 (12.6%)	27 854 (11.1%)	215 (12.6%)
60-69	38 824 (10.1%)	338 (13.5%)	15 224 (10.5%)	237 (9.76%)	27 038 (10.8%)	229 (13.4%)
70-79	73 256 (19.0%)	544 (21.8%)	28 175 (19.4%)	527 (21.7%)	50 006 (19.9%)	388 (22.7%)
80-89	143 644 (37.3%)	875 (35.0%)*	53 534 (36.9%)	911 (37.5%)	89 809 (35.8%)	553 (32.4%)*
90 and above	84 465 (21.9%)	402 (16.1%)***	31 926 (22.0%)	447 (18.4%)***	56 240 (22.4%)	322 (18.9%)**
Sex
Female (ref)	256 180 (66.5%)	1556 (62.3%)	95 517 (65.8%)	1605 (66.1%)	16 6516 (66.4%)	1093 (64.0%)
Male	129 315 (33.5%)	941 (37.7%)***	49 631 (34.2%)	824 (33.9%)	84 431 (33.6%)	614 (36.0%)
Languages
English (ref)	299 717 (77.7%)	1930 (77.3%)	114 960 (79.2%)	1867 (76.9%)	200 083 (79.7%)	1420 (83.2%)
Other	85 778 (22.3%)	567 (22.7%)	30 188 (20.8%)	562 (23.1%)**	50 864 (20.3%)	287 (16.8%)**
Marital status
Single (ref)	197 987 (51.4%)	1236 (49.5%)	72 773 (50.1%)	1150 (47.3%)	125 329 (49.9%)	820 (48.0%)
Partnered	97 888 (25.4%)	641 (25.7%)	34 879 (24.0%)	602 (24.8%)	58 662 (23.4%)	386 (22.6%)
Unknown	89 620 (23.2%)	620 (24.8%)*	37 496 (25.8%)	677 (27.9%)*	66 956 (26.7%)	501 (29.3%)*
Number of contacts on file
0-1 (ref)	96 087 (24.9%)	812 (32.5%)	33 868 (23.3%)	705 (29.0%)	66 404 (26.5%)	537 (31.5%)
2	131 471 (34.1%)	811 (32.5%)***	48 938 (33.7%)	833 (34.3%)***	85 644 (34.1%)	589 (34.5%)**
3 or more	157 937 (41.0%)	874 (35.0%)***	62 342 (43.0%)	891 (36.7%)***	98 899 (39.4%)	581 (34.0%)***
Acuity of care needs
Low risk (ref)	140 155 (36.4%)	886 (35.5%)	51 124 (35.2%)	946 (38.9%)	93 681 (37.3%)	636 (37.3%)
Medium risk	119 807 (31.1%)	761 (30.5%)	46 565 (32.1%)	815 (33.6%)	87 476 (34.9%)	575 (33.7%)
High risk	95 661 (24.8%)	499 (20.0%)**	39 388 (27.1%)	477 (19.6%)***	64 930 (25.9%)	397 (23.3%)
Unknown	29 872 (7.75%)	351 (14.1%)***	8071 (5.6%)	191 (7.9%)*	4860 (1.9%)	99 (5.7%)***
Health region
Toronto Central (ref)	156 805 (40.7%)	1054 (42.2%)	60 403 (41.6%)	1039 (42.8%)	103 099 (41.1%)	731 (42.8%)
Central East	113 341 (29.4%)	827 (33.1%)	41 985 (28.9%)	667 (27.5%)*	73 955 (29.5%)	524 (30.7%)
Central	115 349 (29.9%)	616 (24.7%)***	42 760 (29.5%)	723 (29.8%)	73 893 (29.4%)	452 (26.5%)*
Home care utilization factors
Care setting
Non-congregate (ref)	311 338 (80.8%)	2185 (87.5%)	114 395 (78.8%)	2198 (90.5%)	190 893 (76.1%)	1425 (83.5%)
Congregate	74 157 (19.2%)	312 (12.5%)***	30 753 (21.2%)	231 (9.5%)***	60 054 (23.9%)	282 (16.5%)***
Number of programs
1 (ref)	356 924 (92.6%)	2364 (94.7%)	135 914 (93.6%)	2295 (94.5%)	237 375 (94.6%)	1650 (96.7%)
2 or more	28 571 (7.4%)	133 (5.3%)***	9234 (6.4%)	134 (5.5%)	13 572 (5.41%)	57 (3.34%)***
Care type
ADL (ref)	144 508 (37.5%)	975 (39.0%)	53 069 (36.6%)	1072 (44.1%)	97 615 (38.9%)	683 (40.0%)
ADL + IADL	240 987 (62.5%)	1522 (61.0%)**	92 079 (63.4%)	1357 (55.9%)***	153 332 (61.1%)	1024 (60.0%)
Service intensity	4.7 (4.6)	3.6 (4.0)***	5.0 (4.8)	3.3 (3.6)***	5.0 (4.8)	3.5 (4.1)***
Tenure (weeks)	125.0 (162.6)	81.5 (133.9)***	134.8 (168.0)	100.2 (142.5)***	137.6 (168.5)	83.3 (134.4)***
Tenure
Under 3 months (ref)	81 922 (21.3%)	941 (37.7%)	27 424 (18.9%)	712 (29.3%)	40 243 (16.0%)	663 (38.8%)
3-6 months	32 383 (8.4%)	272 (10.9%)***	11 333 (7.8%)	250 (10.3%)***	23 226 (9.26%)	196 (11.5%)***
6-12 months	52 275 (13.6%)	333 (13.3%)***	17 796 (12.3%)	290 (11.9%)***	33 894 (13.5%)	201 (11.8%)***
12-24 months	71 840 (18.6%)	340 (13.6%)***	27 678 (19.1%)	415 (17.1%)***	45 234 (18.0%)	209 (12.2%)***
Over 24 months	147 075 (38.2%)	611 (24.5%)***	60 917 (42.0%)	762 (31.4%)***	108 350 (43.2%)	438 (25.7%)***
Placed a previous hold
No (ref)	351 533 (91.2%)	1812 (72.6%)	131 314 (90.5%)	1600 (65.9%)	226 617 (90.3%)	958 (56.1%)
Yes	33 962 (8.81%)	685 (27.4%)***	13 834 (9.5%)	829 (34.1%)***	24 330 (9.7%)	749 (43.9%)***
Visit from a new PSW ¹
No (ref)	306 176 (79.4%)	1823 (73.0%)	121 546 (83.7%)	1913 (78.8%)	212 919 (84.8%)	1287 (75.4%)
Yes	79 319 (20.6%)	674 (27.0%)***	23 602 (16.3%)	516 (21.2%)***	38 028 (15.2%)	420 (24.6%)***
Cancellations ¹	0.139 (0.476)	0.294 (0.791)***	0.141 (0.518)	0.381 (1.010)***	0.141 (0.494)	0.404 (0.958)***
Missed cares ¹	0.013 (0.129)	0.020 (0.161)**	0.006 (0.092)	0.005 (0.086)	0.005 (0.093)	0.006 (0.105)

Factors reflect the most recent week that care was received (eg, if on a hold, the last week that care was received included a cancellation).

P < .05, **P < .01, or ***P < .001 for test of comparison between weeks with and without non-hospital service holds, stratified by period (assessed through generalized estimating equations).

During the pre- and early-pandemic periods, clients with more than three contacts on file had similar reductions in the odds of initiating a service hold, which ranged from 22% to 28%; number of contacts on file was not a significant predictor by mid-pandemic (pre-: OR = 0.72, 95% CI [0.62, 0.83]; early-: OR = 0.78, 95% CI [0.67, 0.91]; difference between periods: z_{pre vs early} = non-significant) (Table 3). Compared to those in private homes, clients living in congregate settings had a 41% to 77% reduction in the odds of initiating a hold on homecare services throughout all periods, with a substantial and significant increase in the odds of placing a hold early-pandemic before reverting back by mid-pandemic towards a similar effect size as seen pre-pandemic (pre-: OR = 0.59, 95% CI [0.50, 0.70]; early-: OR = 0.33, 95% CI [0.27, 0.40]; mid-: OR = 0.48, 95% CI [0.39, 0.59]; difference between periods: z_{early vs pre} = 4.43, P < .001; z_{early vs mid} = −2.56, P = .011; z_{pre vs mid} = non-significant). Although less likely to initiate a hold, clients living in congregate settings who did pause their care had significantly shorter holds during the early-pandemic period than those living in non-congregate settings (pre-: non-significant; early-: expβ = 0.79, 95% CI [0.68, 0.92]; mid-: non-significant).

Table 3.

Predictors of homecare service hold initiation and length, stratified by period.

	Hold initiation			Hold length
Client characteristics	Pre-pandemic	Early-pandemic	Mid-pandemic	Pre-pandemic	Early-pandemic	Mid-pandemic
Client characteristics	OR (95% CI)			expβ (95% CI)
Demographics
Age (years )
Under 60	—	—	—	—	—	—
60-69	1.02 (0.82, 1.28)	0.67 (0.52, 0.87)**	0.87 (0.64, 1.18)	0.86 (0.72, 1.01)	0.84 (0.70, 1.02)	0.85 (0.68, 1.05)
70-79	0.85 (0.70, 1.04)	0.90 (0.72, 1.12)	0.92 (0.70, 1.22)	0.79 (0.68, 0.93)**	1.08 (0.92, 1.27)	0.97 (0.80, 1.18)
80-89	0.81 (0.67, 0.97)*	0.88 (0.71, 1.08)	0.86 (0.66, 1.13)	0.83 (0.72, 0.96)*	0.99 (0.86, 1.15)	0.97 (0.81, 1.17)
90 and above	0.69 (0.56, 0.86)**	0.86 (0.69, 1.08)	1.06 (0.80, 1.41)	0.80 (0.68, 0.94)**	1.19 (1.01, 1.40)*	0.88 (0.72, 1.08)
Sex
Female	—	—	—	—	—	—
Male	1.09 (0.97, 1.23)	0.88 (0.77, 0.999)*	1.00 (0.85, 1.17)	0.96 (0.87, 1.05)	0.91 (0.83, 1.00)	1.04 (0.93, 1.17)
Languages
English	—	—	—	—	—	—
Other	1.09 (0.95, 1.25)	1.25 (1.08, 1.44)**	0.95 (0.78, 1.16)	1.03 (0.92, 1.14)	1.02 (0.92, 1.13)	1.10 (0.95, 1.28)
Marital status
Single	—	—	—	—	—	—
Partnered	0.94 (0.81, 1.08)	1.00 (0.85, 1.16)	0.96 (0.78, 1.17)	1.11 (0.99, 1.24)	1.11 (0.99, 1.24)	1.01 (0.87, 1.16)
Unknown	0.97 (0.85, 1.12)	0.99 (0.86, 1.15)	1.08 (0.90, 1.30)	0.98 (0.88, 1.10)	1.09 (0.98, 1.22)	1.04 (0.92, 1.19)
Number of contacts on file
0-1	—	—	—	—	—	—
2	0.79 (0.69, 0.91)**	0.89 (0.76, 1.03)	1.06 (0.87, 1.28)	1.02 (0.91, 1.13)	0.98 (0.87, 1.09)	1.03 (0.90, 1.18)
3 or more	0.72 (0.62, 0.83)***	0.78 (0.67, 0.91)**	1.01 (0.83, 1.24)	0.97 (0.87, 1.09)	0.92 (0.82, 1.03)	1.00 (0.86, 1.16)
Acuity of care needs
Low risk	—	—	—	—	—	—
Medium risk	1.01 (0.88, 1.17)	0.84 (0.73, 0.97)*	0.86 (0.71, 1.03)	1.03 (0.92, 1.15)	0.96 (0.86, 1.06)	0.88 (0.77, 1.00)
High risk	0.83 (0.70, 0.98)*	0.59 (0.50, 0.71)***	0.86 (0.69, 1.07)	0.92 (0.81, 1.05)	0.88 (0.77, 1.00)	0.84 (0.72, 0.99)*
Unknown	1.76 (1.47, 2.12)***	0.89 (0.70, 1.12)	1.64 (1.17, 2.31)**	0.86 (0.75, 0.99)*	0.59 (0.49, 0.70)***	0.66 (0.52, 0.84)**
Health region
Toronto Central	—	—	—	—	—	—
Central East	1.01 (0.87, 1.16)	0.83 (0.71, 0.98)*	0.98 (0.81, 1.19)	0.89 (0.80, 0.995)*	1.41 (1.25, 1.58)***	1.02 (0.89, 1.17)
Central	0.72 (0.62, 0.83)***	1.02 (0.88, 1.18)	0.88 (0.73, 1.07)	1.02 (0.91, 1.14)	2.24 (2.01, 2.50)***	0.95 (0.82, 1.09)
Homecare utilization factors
Care setting
Non-congregate	—	—	—	—	—	—
Congregate	0.59 (0.50, 0.70)***	0.33 (0.28, 0.40)***	0.48 (0.39, 0.59)***	0.94 (0.82, 1.07)	0.79 (0.68, 0.92)**	1.12 (0.96, 1.31)
Number of programs
1	—	—	—	—	—	—
2 or more	0.74 (0.57, 0.96)*	1.00 (0.75, 1.32)	0.60 (0.38, 0.94)*	0.84 (0.69, 1.03)	0.99 (0.81, 1.21)	0.82 (0.60, 1.13)
Care type
ADL	—	—	—	—	—	—
ADL + IADL	0.90 (0.80, 1.01)	0.61 (0.54, 0.69)***	0.79 (0.67, 0.93)**	1.00 (0.91, 1.10)	0.88 (0.80, 0.96)**	0.90 (0.80, 1.02)
Service intensity	0.97 (0.96, 0.98)***	0.93 (0.91, 0.94)***	0.94 (0.92, 0.96)***	0.97 (0.96, 0.98)***	0.99 (0.97, 0.998)*	0.96 (0.94, 0.97)***
Tenure
Under 3 months	—	—	—	—	—	—
3-6 months	0.71 (0.61, 0.82)***	0.83 (0.69, 0.999)*	0.54 (0.44, 0.66)***	1.00 (0.87, 1.16)	1.50 (1.28, 1.77)***	1.21 (0.99, 1.49)
6-12 months	0.51 (0.44, 0.59)***	0.52 (0.43, 0.63)***	0.34 (0.27, 0.42)***	1.10 (0.95, 1.26)	1.47 (1.25, 1.72)***	1.18 (0.97, 1.43)
12-24 months	0.39 (0.33, 0.45)***	0.41 (0.34, 0.49)***	0.16 (0.13, 0.20)***	1.11 (0.96, 1.27)	1.47 (1.27, 1.70)***	1.18 (0.97, 1.43)
Over 24 months	0.32 (0.28, 0.37)***	0.37 (0.31, 0.44)***	0.12 (0.09, 0.14)***	1.09 (0.96, 1.24)	1.38 (1.21, 1.58)***	1.07 (0.91, 1.27)
Placed a previous hold
No	—	—	—	—	—	—
Yes	6.10 (5.28, 7.04)***	13.02 (11.07, 15.30)***	39.86 (31.81, 49.95)***	0.93 (0.85, 1.02)	0.79 (0.72, 0.87)***	1.07 (0.95, 1.21)
Visit from a new PSW ¹
No	—	—	—
Yes	1.11 (1.01, 1.23)*	1.24 (1.11, 1.39)***	1.13 (0.99, 1.28)
Cancellations ¹	1.35 (1.27, 1.43)***	1.57 (1.49, 1.66)***	1.60 (1.50, 1.70)***
Missed cares ¹	1.03 (0.80, 1.32)	0.57 (0.35, 0.93)*	0.86 (0.56, 1.31)
Marginal R²/Conditional R²	.111/.489	.191/.530	.221/.635	.034/.340	.143/.250	.048/.328
ICC	0.425	0.419	0.531	0.294	0.125	0.317

Factors reflect the most recent week that care was received (eg, if on a hold, the last week that care was received included a cancellation).

P < .05, **P < .01, or ***P < .001.

The degree to which acuity of care needs influenced decisions to pause homecare varied during the pandemic. The reduction in the odds of pausing homecare services for clients with the highest care acuity doubled from 17% odds of pausing care pre-pandemic to 41% by early-pandemic before becoming non-significant by mid-pandemic (pre-: OR = 0.83, 95% CI [0.70, 0.98]; early-: OR = 0.59, 95% CI [0.50, 0.71]; mid-: non-significant; difference between periods: z_{early vs pre} = 2.65, P = .008). For clients who placed a service hold mid-pandemic, those with higher acuity paused services for a shorter time span than those with low acuity (expβ = 0.84, 95% CI [0.72, 0.99]). Receiving care for both IADLs and ADLs (as opposed to ADLs only), significantly predicted lower odds of hold initiation in both early- and mid-pandemic periods; the reduction in odds was significantly lower during the early- versus mid-pandemic period (pre-: non-significant; early-: OR = 0.61, 95% CI [0.54, 0.69]; mid-: OR = 0.79, 95% CI [0.67, 0.93]; difference between periods: z_{early vs mid} = −2.49, P = .013).

Clients who received a greater number of personal support service types (ie, personal care, homemaking, respite services) had significantly lower odds of initiating service holds in both the pre- and mid-pandemic periods (pre-: OR = 0.74, 95% CI [0.57, 0.96]; early-: non-significant; mid-: OR = 0.60, 95% CI [0.38, 0.94]; difference between periods: z_{pre vs mid} = non-significant). Clients with higher service intensity had slightly lower odds of initiating a hold across all periods; this effect was significantly less pronounced pre- versus early- or mid-pandemic (pre-: OR = 0.97, 95% CI [0.96, 0.98]; early-: OR = 0.93, 95% CI [0.91, 0.94]; mid-: OR = 0.94, 95% CI [0.92, 0.96]; difference between periods: z_{pre vs early} = 3.90, P < .001; z_{pre vs mid} = 2.66, P < .001; z_{early vs mid} = non-significant). Amongst clients who had placed a hold, those with higher service intensity tended to pause their services for a shorter period than those with lower service intensity, although the effect size was small for all models and significantly smaller early-pandemic versus mid-pandemic (pre-: expβ = 0.97, 95% CI [0.96, 0.98]; early-: expβ = 0.99, 95% CI [0.97, 0.998]; mid-: expβ = 0.96, 95% CI [0.94, 0.97]; differences between periods: z_{pre vs early} = non-significant; z_{pre vs mid} = non-significant; z_{early vs mid} = 2.91, P < .001).

Clients with longer tenures as homecare recipients had a 17% to 88% reduction in the odds of initiating a service hold compared to clients with the shortest tenure (<3 months) across all models. The differences in effect size of longer tenure were significantly greater by the mid-pandemic period, with tenure longer than 2 years resulting in 88% reduced odds of initiating a hold mid-pandemic versus 68% pre-pandemic and 63% early-pandemic (pre-: OR = 0.32, 95% CI [0.28, 0.37]; early-: OR = 0.37, 95% CI [0.31, 0.44]; mid-: OR = 0.12, 95% CI [0.09, 0.14]; difference between periods: z_{pre vs early} = non-significant; z_{pre vs mid} = 7.58, P < .001; z_{early vs mid} = 8.19, P < .001).

Receiving care from a new PSW correlated with an 11% to 24% increase in the odds of initiating a hold within the first two periods and the strength of the effect was not significantly different across models (pre-: OR = 1.11, 95% CI [1.01, 1.23]; early-: OR = 1.24, 95% CI [1.11, 1.39]; mid-: non-significant).

Past decisions to pause or cancel homecare services substantially increased the likelihood of initiating further holds (holds: (pre-: OR = 6.10, 95% CI [5.28, 7.04]; early-: OR = 13.02, 95% CI [11.07, 15.30]; mid-: OR = 39.86, 95% CI [31.81, 49.95]); cancellations: (pre-: OR = 1.35, 95% CI [1.27, 1.43]; early-: OR = 1.57, 95% CI [1.49, 1.66]; mid-: OR = 1.60, 95% CI [1.50, 1.70]). The effect of a past cancellation on the decision to pause care was significantly stronger during the pandemic periods versus pre-pandemic (difference between periods: z_{pre vs early} = −3.82, P < .001; z_{pre vs mid} = −3.96, P < .001; z_{early vs mid} = non-significant), while the effect of a previous hold increased significantly over time between each sequential model (difference between periods: z_{pre vs early} = −6.86, P < .001; z_{pre vs mid} = −13.74, P < .001; z_{early vs mid} = −7.90, P < .001). Experiencing missed care (eg, visit missed due to PSW calling in sick) was significantly less frequent during the pandemic periods (Table 1); in the early-pandemic period, clients who did experience missed care had significantly decreased odds of placing their services on hold (pre-: non-significant; early-: OR = 0.57, 95% CI [0.35, 0.93]; mid-: non-significant).

Discussion

In this paper, we examined the relative effects of individual-level factors on the odds of placing a hold on homecare services and the length of these homecare service holds during pre-, early-, and mid-pandemic periods. Our regression analyses, which utilized a large longitudinal administrative dataset, expanded on the otherwise largely qualitative and descriptive literature.^2,3,19 Findings revealed the relative significance of care needs on homecare receipt, where care needs were indicated by service intensity, acuity of care needs, comprehensive services (eg, both ADL and IADL care), and tenure as a homecare client. Strengthening knowledge in this area is particularly relevant given that the number and length of holds on homecare services more than doubled during the period examined (ie, pre- to early-pandemic; Figures 1 and 2), with previously-documented consequences of reduced home care support for clients and their families.^5,7,8,9 We found that the return to pre-pandemic hold rates by mid-pandemic (as shown in Figure 1) masked changes in the individual-level drivers of these choices which were revealed through the stratified regression models. Although there was some variation across models, clients with higher care needs generally had significantly lower odds of initiating a hold on homecare services. Past choices to forgo homecare, expressed through cancellations of single visits or through placing previous holds, emerged as the strongest predictors of clients’ decisions to pause home care services.

Figure 2.

Average duration of client-initiated holds over time.

Care Needs

Clients with higher care needs may be more dependent on homecare services, resulting in a lower likelihood of initiating a hold early-pandemic. The influence of care needs on initiating holds on homecare services can be seen through multiple variables, including lower odds of pausing care for clients who: received ADL and IADL services (significant across all models), received multiple types of personal support services (significant pre- and mid-pandemic), and had high care acuity (significant pre- and early-pandemic). Although, once accounting for other factors, the effect size of service intensity was relatively minor, clients with comparatively lower service intensity who did initiate a hold early-pandemic tended to do so for shorter periods of time. This likely reflects the difficult decisions that clients with lower care requirements and their families were making as they balanced the fear of infection against potential detriment to both client and caregiver wellbeing from a reduction in these homecare services.^5,8,9,20 In future research, it will be important to uncover specific barriers to receiving care as well as potential health ramifications for lower need clients who choose to forgo homecare services.

Support Network

Living in congregate settings (eg, retirement homes) or having a wider documented support system (greater number of contacts on file with the care organization) could signal lower independence, and therefore, may also relate to these clients’ reduced likelihood of pausing homecare services. Although based on previous literature, we had expected those with higher social support to have lower homecare utilization,^5,11,19 our analyses showed that having a greater number of contacts on file (the best available proxy for social support) had a protective effect against placing a hold on homecare pre- and early-pandemic, though this effect was non-significant in the mid-pandemic period. Although significantly less likely to have initiated a hold early-pandemic, clients who lived in congregate settings who did pause their homecare services at the onset of the pandemic did so for a significantly shorter period than those in non-congregate settings. Clients in congregate settings who briefly paused care during early- and mid-pandemic periods may have done so out of necessity, due to some retirement homes temporarily closing access to visitors – including homecare workers – due to heightened infection control measures.²¹ There may also have been an expectation that congregate setting staff or family members could fill in the gaps in care, before the limitations in the ability to stretch these alternative care resources became clear (eg, due to staffing capacity issues in congregate settings²² and visit restrictions for family caregivers²³).

Past Homecare Experiences

Previous experiences with homecare services were significantly linked to future decisions around homecare receipt. Clients with longer tenures as homecare clients were less likely to place a hold across all examined periods, even in the regression models which controlled for factors such as client age, acuity and social support (Table 3). Clients with longer tenure may have become accustomed to relying on homecare and/or had a stronger bond and sense of trust with their long-term PSWs leading to a higher likelihood of choosing to continue receiving homecare services. The importance of these relationships between client and PSW is reinforced by the finding that holds were more likely (11-24%) to be placed by clients who had received care from a new PSW in the previous week during pre- and early-pandemic periods. Previous qualitative work has found that receiving care from multiple PSWs contributed to client and family decisions to forgo home care at the onset of the pandemic.⁹ Although we also found that provider consistency contributed to decisions to pause care during the early-pandemic period, this did not emerge as a major driver of increased holds relative to other factors in our models.

Ability to Forgo Care

Beyond the influence of social support and acuity, ability to forgo care is also reflected in clients’ previous care decisions, expressed through cancelling individual visits or placing a service hold. These were both strong significant predictors of clients’ decisions to pause home care services during all periods, with significantly more pronounced effects during the pandemic. Previous holds were by far the strongest predictor of placing another hold during each of the three time periods studied. By contrast, in the rare cases where clients went without care involuntarily due to the organization being unable to fulfil a care visit (averaging 0.009 missed care visits in the most recent week care was received across the entire study period), the early-pandemic likelihood of placing a service hold decreased. While it is perhaps unsurprising that individuals who had previously made the decision to go without care would be more likely to do so again, these associations have not previously been reported. The present findings highlight the value of incorporating clients’ past decisions to forgo home care services in predicting future decision-making.

Limitations

Although the study utilized a large sample with a wide array of variables, the data came from a single homecare organization’s administrative data that was initially captured for service planning rather than research purposes. As such, data was not always complete, and any services provided by other organizations or purchased privately would not be visible. The aim of this study was to examine the decisions of clients to pause their publicly-funded homecare services; there may be differences in decision-making by clients paying privately for their care which are not captured in this paper. As the administrative data did not always include all factors of interest, some proxies were used. For instance, to address the lack of available information related to race, ethnicity, or national identity, “language spoken” was the closest proxy available and was therefore included to control for this limitation to the extent possible. Similarly, while the potential extent of family caregiver support was not directly captured in the available data, the number of contacts on file has previously demonstrated a significant relationship to homecare outcomes²⁴ and was used as a proxy for the extent of available support.

Conclusions

Findings presented in this paper provide insights for service planning by highlighting characteristics and experiences of individuals who chose to forgo homecare services over time. The number of service holds placed more than doubled in the early-pandemic period, before shrinking back towards pre-pandemic rates in the mid-pandemic period. Although basic trends generally indicated a return to pre-pandemic levels by the mid-pandemic period, our regression analyses highlighted important distinctions following from the pandemic that did not return to “normal.” Despite differences in the degree of influence across time periods, a general finding was that clients with higher care needs were less likely to place a hold on their homecare services. Prior homecare experiences, including shorter tenure as a homecare client and previously choosing to pause homecare, were also a good indication of future decisions to place a hold on homecare services. These results provide a better understanding of the care receipt decisions made by a large sample of homecare clients and can be used by care delivery organizations, policymakers, and those interested in forecasting to inform resource planning and guide prioritization of service delivery during capacity crises.

Footnotes

ORCID iDs

Prakathesh Rabeenthira

Katherine A. P. Zagrodney

Emily C. King

Kathryn A. Nichol

Sandra M. McKay

References

Organisation for Economic Co-operation and Development. Health at a Glance 2023: OECD Indicators. Paris: OECD Publishing, 2003.

Sinn

Sultan

Turcotte

McArthur

Hirdes

JP.

Patterns of home care assessment and service provision before and during the COVID-19 pandemic in Ontario, Canada. PLoS One. 2022;17(3):e0266160.

Jones

Maclagan

Schumacher

, et al. Impact of the COVID-19 pandemic on home care services among community-dwelling adults with dementia. J Am Med Dir Assoc. 2021;22(11):2258-2262.e1.

Sama

Quinn

Galligan

, et al. Impacts of the COVID-19 pandemic on home health and home care agency managers, clients, and aides: a cross-sectional survey, March to June, 2020. Home Health Care Manag Pract. 2021;33(2):125-129.

van den Bulck

AOE

de Korte

Metzelthin

, et al. In the eye of the storm: a quantitative and qualitative account of the impact of the COVID-19 pandemic on Dutch home healthcare. Int J Environ Res Public Health. 2022;19(4):2252.

King

Zagrodney

Rabeenthira

Van Belle

McKay

SM.

Why did home care personal support service volumes drop during the COVID-19 pandemic? The contributions of client choice and personal support worker availability. Health Serv Insights. 2023;16:1-6.

Kim

Yee

, et al. “I depend on her for everything”: a retrospective chart review of home care worker service disruptions for homebound older adults during the COVID-19 pandemic. J Appl Gerontol. 2023;42(4):561-570.

Reckrey

Perez

Watman

Ornstein

Russell

Franzosa

(2023). The need for stability in paid dementia care: family caregiver perspectives. J Appl Gerontol. 2023;42(4):607-616.

Weeks

Nesto

Hiebert

, et al. Health service experiences and preferences of frail home care clients and their family and friend caregivers during the COVID-19 pandemic. BMC Res Notes. 2021;14(1):271.

10.

Nizzer

Ruco

Moreira

, et al. “You have to be careful about every detail” How the COVID-19 pandemic shaped the experiences of Canadian personal support workers working in home care. J Occup Environ Med. 2023;65(9):e604-e609.

11.

Vislapuu

Angeles

Berge

Kjerstad

Gedde

Husebo

. The consequences of COVID-19 lockdown for formal and informal resource utilization among home-dwelling people with dementia: results from the prospective PAN.DEM study. BMC Health Serv Res. 2021;21(1):1003.

12.

Giebel

Hanna

Cannon

, et al. Decision-making for receiving paid home care for dementia in the time of COVID-19: a qualitative study. BMC Geriatr. 2020;20:1-8.

13.

Ontario Agency for Health Protection and Promotion (Public Health Ontario). Weekly Epidemiologic Summary: COVID-19 in Ontario – September 18, 2022 to September 24, 2022. Toronto, ON: King’s Printer for Ontario, 2022. Accessed April 19, 2024. https://www.publichealthontario.ca/-/media/Documents/nCoV/epi/2022/09/covid-19-weekly-epi-summary-report-sept-30.pdf?rev=05a73172a54d4172b83792224ecb5caf&sc_lang=en

14.

Zeger

Liang

Albert

PS.

Models for longitudinal data: a generalized estimating equation approach. Biometrics. 1988;44(4):1049-1060.

15.

Hanley

Negassa

deB. Edwardes

Forrester

JE.

Statistical analysis of correlated data using genralized estimating equations: an orientation. Am J Epidemiol. 2003;157(4):364-375.

16.

Clogg

Petkova

Haritou

Statistical methods for comparing regression coefficients between models. Am J Sociol. 1995;100(5):1261-1293.

17.

Snee

RD.

Regression diagnostics: identifying influential data and sources of collinearity. J Qual Technol. 1983;15(3):149-153.

18.

Canadian Institutes of Health Research, Natural Sciences and Engineering Research Council of Canada, and Social Sciences and Humanities Research Council. Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans. Ottawa, ON: Government of Canada, 2018. Accessed March 17, 2025. https://ethics.gc.ca/eng/documents/tcps2-2018-en-interactive-final.pdf

19.

Mah

Stevens

Keefe

Rockwood

Andrew

MK.

Social factors influencing utilization of home care in community-dwelling older adults: a scoping review. BMC Geriatr. 2021;21(145):1-21.

20.

Bergmann

Wagner

The impact of COVID-19 on informal caregivers in the US. Int J Aging Res. 2021;4(3):87.

21.

Ontario Ministry of Health, Officer of the Chief Medical Officer of Health. COVID-19 Directive #3 for Long-Term Care Homes Under the Long-Term Care Homes Act, 2007. Issued under Section 77.7 of the Health Protection and Promotion Act (HPPA), R.S.O. 1990, c. H.7. 2020. Toronto, ON: Government of Ontario, 2020. https://www.publications.gov.on.ca/CL30081

22.

Titley

Young

Savage

Thorne

Spiers

Estabrooks

CA.

Cracks in the foundation: the experience of care aides in long-term care homes during the COVID-19 pandemic. J Am Geriatr Soc. 2023;71(1):198-205.

23.

Chu

Yee

Stamatopoulos

“It’s the worst thing I’ve ever been put through in my life”: the trauma experienced by essential family caregivers of loved ones in long-term care during the COVID-19 pandemic in Canada. Int J Qual Stud Health Well-Being. 2022;17(1):2075532.

24.

Saragosa

Zagrodney

KAP

Rabeenthira

King

McKay

SM.

How might we have known? Using administrative data to predict 30-day hospital readmission in clients receiving home care services from 2018 to 2021. Health Serv Insights, 2023;16:1-8.