This was a secondary analysis of data from the COVID-19 and the Care of Assisted living Residents in Alberta (COVCARES-AB) study. We linked surveys from a key contact in each participating AL home to administrative resident data in these homes. The survey was based on a validated instrument used in a prospective AL and NH cohort study [34, 35] and included items derived from the US national study of AL [36], with additional ones developed by the study team. For COVCARES-AB, we added items related to COVID-19 [37, 38] modified for the Canadian AL context. Routinely collected, de-identified Resident Assessment Instrument – Home Care (RAI-HC) [39, 40] records provided by the Alberta Health Services (AHS) Enterprise Data Warehouse were used to characterize residents.
Setting and sample
All AL homes licensed and publicly subsidized (called Designated Supportive Living) [41] by AHS, were considered. We excluded homes that (a) were in operation for < 6 months, (b) primarily served psychiatric residents, or (c) had < 4 residents 65 years or older in smaller homes (below median bed-size in their health zone) or < 10 residents in larger homes (above median bed-size).
We invited a key contact from each of the 163 eligible AL homes to participate in our survey. Key contacts were individuals anticipated to be familiar with the day-to-day operation of the home (e.g., home administrators, directors of care, care managers). An initial email was sent, followed by up to 3 email reminders at approximately 2-week intervals. We contacted homes by phone that did not respond to our emails. From October 26, 2020 to April 7, 2021, we collected surveys from 64 AL homes (response rate: 39%). We inquired about the homes' experiences in the three months before and after the start of visitor restrictions in congregate care in Alberta (March 1, 2020) [42]. In Alberta, pandemic wave 1 spanned March 1, 2020 to June 30, 2020 [43]. We subsequently invited all 64 participating AL homes to complete a second survey focusing on experiences during pandemic wave 2 (October 1, 2020 to February 28, 2021) [43]. From July 10 to September 17, 2021, 44 of the 64 homes (69%) responded.
Wave 1 and 2 resident data were extracted from RAI-HC records for all individuals residing in one of the 44 AL homes that submitted two surveys. The RAI-HC is completed on each AL resident in Alberta at admission and then annually or when resident status changes significantly [40]. We excluded 2 homes where no RAI-HC records were available for either of the 2 time periods. Of our 42 included homes, 41 had RAI-HC data from both periods, and one had RAI-HC data only from wave 2. If a resident had more than one RAI-HC assessment in a period (e.g., because of a significant status change after a regular assessment), we used the latest assessment.
Study variables
Dependent variable – pain
In accordance with RAI-HC guidelines [44], we treated pain as a dichotomous variable. Its presence was defined as either daily pain (RAI-HC item k4a ≥ 2) of at least moderate intensity (RAI-HC item k4b ≥ 2) or pain of at least severe intensity (k4b ≥ 3) occurring at least once in the previous week. While these items have high reliability and validity in general, they underestimate pain among persons with cognitive impairment [39, 45], Staff factors like shortages can affect timeliness and accuracy of assessments [46–48]. NHs and AL homes in Canada are required to routinely assess resident pain (including during the COVID-19 pandemic). This information is publicly reported as a national quality of care indicator [49, 50]. Others have used these items to assess pandemic-related changes in NH resident pain [14].
Independent variables – COVID-19-specific contextual conditions in assisted living
Nine variables from our facility survey (Table 1) were considered. Selection was based on evidence suggesting a possible association with resident pain. Facility variables were linked to each resident in the respective AL home, using unique resident and AL home identifiers.
Table 1
Facility survey variables included in the analyses
Construct | Survey question | Response options | Study variable | Hypotheses |
Preparedness for COVID-19 outbreaks | How well prepared was your home for a COVID-19 outbreak? | 5-point Likert scale (1 = very well prepared to 5 = not at all prepared) | Dichotomous (very well or well prepared vs somewhat, poorly, or not at all prepared) | Better preparedness associated with lower odd of pain |
Availability of a registered nurse | Did you have 24-hour registered nurses on-site? | Yes/no | Dichotomous variable | Availability of a registered nurse associated with lower odd of pain |
Availability of a nurse practitioner | Was there a nurse practitioner affiliated (employed/contracted) with your home? | Yes/no | Dichotomous variable | Availability of a nurse practitioner associated with lower odd of pain |
Direct care staff shortage | Did your home experience challenges with ensuring care aide staffing levels? | 5-point Likert scale (1 = no staff shortage to 5 = not available at all) | Dichotomous (moderate or higher shortage of either care aide staffing or licensed practical nurse staffing vs no or limited shortage of care aides and licensed practical nurses) | Direct care staff shortage associated with higher odd of pain |
Did your home experience challenges with ensuring licensed practical nurse staffing levels? | 5-point Likert scale (1 = no staff shortage to 5 = not available at all) |
Decreased staff morale | Was there a change in the morale among staff in the home? | 5-point Likert scale (1 = much worse to 5 = much improved) | Dichotomous (somewhat or much worse vs no change, somewhat, or much improved) | Decreased staff morale associated with higher odd of pain |
COVID-19 outbreaks | Did your home experience an outbreak of COVID-19? | Yes/no | Dichotomous variable | Outbreaks associated with higher odd of pain |
Residents required to remain in their rooms | Was the following measures implemented in your home: residents asked to remain in their rooms? | Yes/no | Dichotomous variable | Residents required to remain in their rooms associated with higher odd of pain |
Video calls with physicians | Did your home implement video calling with residents’ physician(s)? | Yes/no | Dichotomous variable | Availability of vide calls with physicians associated with lower odd of pain |
Caregiver involvement | Do you believe your home created or facilitated opportunities for family/friend caregivers to be well informed and involved in the care and well-being of residents? | Yes/no | Dichotomous variable | Caregiver involvement associated with lower odd of pain |
Preparedness for COVID-19 outbreaks
Facilities with better perceived pandemic preparedness had less disruption of care processes and staff quality of work-life from public health restrictions [31–33].
Availability of registered nurses on-site
Staff knowledge and skills in pain assessment and management are critical in addressing pain [2, 4, 5]. RNs are trained in clinical assessments, and appropriate pain management is within their scope of practice [51, 52]. RNs are not present in person at all times in AL and may only be on call.
Availability of a nurse practitioner
NPs are RNs with additional training and experience that allows them to autonomously diagnose and treat health conditions. They have advanced training in detecting and managing pain [53]. They were crucial in managing chronic conditions and related symptoms in congregate care during the COVID-19 pandemic [54].
Direct care staff shortage
Care aides (personal support workers, nursing assistants) [55] and licensed practical nurses (LPNs) [56] provide direct care in AL. Care aides are unregulated workers responsible for the majority of direct care in Canadian AL homes [55]. LPNs have graduated from a two-year college level course in practical nursing. Their scope of practice is more foundational and focused than that of RNs [56]. Staffing shortages are associated with higher pain prevalence [2, 4, 26]. Restricting staff from working in more than one facility and COVID-19 infections among staff may have led to more severe staffing shortages than before the pandemic [57].
Decreased staff morale
COVID-19 negatively affected staff morale [58]. Moral distress can negatively affect care practices [59].
COVID-19 outbreaks
Despite infection prevention and control protocols [60], COVID-19 outbreaks were still common in AL, especially during wave 2 of the pandemic [61]. Outbreaks may have exacerbated the issues described above. For example, isolating infected residents would lead to additional work for staff, COVID-19 infections among staff would have increased staffing shortages, and/or regular therapies and activities may have been more disrupted in facilities with an outbreak [60].
Confinement of residents to their rooms
Confining residents to their rooms may have negatively affected residents’ mobility, propensity to falling, and mental health [62], all important risk factors for pain [2, 4, 5]. Staff may have seen residents confined to their rooms less frequently and for shorter durations, further reducing opportunities to detect and act on pain.
Video calls with physicians
Access to a physician is critical for adequate and timely treatment of pain [2, 4, 5]. Public health restrictions limited NH and AL residents’ access to physicians. Transfers to health services external to the home were limited to emergencies or essential treatments, and physicians were often reluctant to visit residents in the home [63–65]. Video calls with physicians were often the only option available for residents to see and interact with a physician.
Caregiver involvement
Family/friend caregivers play a crucial role in caring for older adults. Due to the lower staffing and fewer services provided in AL than NHs, caregiver involvement in resident care is even more critical in AL [66–70]. Caregivers can detect symptoms (such as pain) to be addressed by staff, advocate on behalf of the resident for treatment, or intervene to mitigate pain. Visitor restrictions can impede caregivers from fulfilling these essential roles.
Covariates
Based on known resident-level risk factors for pain [2, 4, 5]. we adjusted for the following resident covariates obtained from RAI-HC records: sex, significant physical impairment (Activities of Daily Living Hierarchy [71] score ≥ 3), significant cognitive impairment (RAI-HC Cognitive Performance Scale [72] score ≥ 3), significant depressive symptoms (RAI-HC Depression Rating Scale [73] score ≥ 3), and health instability (RAI-HC Changes in Health, End-Stage Disease, Signs, and Symptoms Scale [74] score ≥ 2). These scales aggregate various co-morbid conditions, symptoms and functional variables, increasing parsimony of our regression models (compared to including all these variables individually).
Statistical analyses
SAS® 9.4 was used for all analyses. The resident sample was described by frequency and percent distributions of all included variables by time of data collection (overall and by pain status). The facility sample was also described by frequency distributions for all variables at the time of data collection. We compared outcomes between pandemic waves 1 and 2 using general estimating equations with each of the outcomes as dependent variables, wave (1 vs 2) as an independent variable, and a random term to account for repeated measures within residents and facilities. Using Chi2 tests or Fisher’s exact tests, we compared outcomes between residents with and without moderate daily or severe pain at any frequency within each wave.
We assessed factors associated with resident pain during the pandemic using general estimating equations, accounting for repeated measures within residents and with a robust sandwich estimator to account for non-normality and clustering of residents within facilities. Our dependent variable was daily or at least severe pain of any frequency, and we added time of assessment (wave 1 vs 2) as a covariate. In a blocked stepwise approach, we first included all facility survey variables with a p value of ≤ 0.15 in our bivariate analyses comparing residents with and without pain. We added variables one-by-one and removed those that caused collinearity issues or negatively affected model fit (increased − 2 Res Log Pseudo-Likelihood and/or generalized Chi-square). Next, we included resident covariates, using the same approach, and finally added the health zone covariate. We considered adjusting for resident age, facility-level average pre-pandemic (January 1, 2018 to February 29, 2020) pain prevalence, and new admissions, but these factors were not significantly associated with resident pain and were not included in our final models. We also considered adjusting for facility-level sampling strata (region, bed-size, ownership, urban vs rural location). These variables were highly intercorrelated, and we found that adding them simultaneously did not improve explanatory power and decreased model fit. We adjusted for health zone, the only facility-level sampling variable significantly associated with resident pain.
To determine whether the measurement of pain had systematically changed after the onset of the pandemic or between residents with versus without cognitive impairment, we assessed the pain variable for differential item functioning (DIF), using well-established methods [75, 76] (details in Additional File 1). We found no evidence of DIF.
To assess the representativeness of our sample, we compared characteristics of included and excluded facilities and residents. Excluded facilities (i.e., those that did not submit two facility surveys) were more likely to be small (< 40 beds) and located in the Calgary and South zones (Additional File 2). Residents in excluded homes (Additional File 2) were more likely to be > 85 years old, have post-secondary education and health instability, live in homes in the Calgary or South zones, of small size, with a private for-profit model, and in rural areas (wave 1 only). Residents in included homes were more likely to have dementia.
We used multiple imputations to account for missing data. There were no missing data for resident variables. There were missing data on facility preparedness and availability of an RN or NP for one facility (wave 1), family involvement variable for one facility (wave 1), and information on COVID-19 outbreaks for 3 facilities (wave 2).