key: cord-0702737-1cv47ezz authors: Veldhuizen, Scott; Mahinpey, Newsha; Zawertailo, Laurie; Minian, Nadia; Melamed, Osnat; Selby, Peter title: Effects of COVID-19-related disruptions on service use in a large smoking-cessation program date: 2022-04-14 journal: Nicotine Tob Res DOI: 10.1093/ntr/ntac103 sha: 7634f16ba60ab60e8ffaece92f39b4f418fe6722 doc_id: 702737 cord_uid: 1cv47ezz BACKGROUND: The COVID-19 pandemic caused a rapid shift to virtual care, with largely unknown consequences for accessibility. The purpose of this study is to examine pandemic-related changes in use of care for smoking cessation. METHODS: We conducted a secondary analysis 65,565 enrolments in a large smoking cessation program in Ontario, Canada. We used piecewise mixed effects regression to examine how weeks NRT received, as well as counseling provided and visits attended, varied with date of enrolment over three time periods: more than 6M before the pandemic began; the 6M before the pandemic; and the pandemic period itself. We then examined changes in the associations between use of care and participant characteristics by fitting a model including a set of interactions between time and other variables. Based on an omnibus test of these interactions, we then tested individual terms, using the Holm method to control the familywise error rate. RESULTS: From the start of the pandemic in March, 2020, total weeks of NRT provided rose significantly and then declined, while the amount of counseling fell. Associations between NRT use and participant characteristics changed significantly after the pandemic onset. Individual models showed that people with lower income, living in areas of higher marginalization, unable to work, and reporting higher levels of depressive symptoms all received NRT for a longer time during the pandemic period. CONCLUSION: The pandemic led to small but significant changes in the amount of services used per enrolment. The transition to remote care appears to have reduced the effects of socioeconomic and health barriers. IMPLICATIONS: The amount of care used by participants in tobacco cessation treatment is known to be associated with health and sociodemographic characteristics. Most of these associations did not change markedly following the pandemic-related switch to virtual care in 2020; however, the effects of some economic and health barriers seems to have lessened, perhaps due to a likely reduction in transport and time requirements of treatment. The World Health Organization, alongside other bodies, has called for tobacco cessation to be treated as a priority during the COVID-19 pandemic 1-3 . Some evidence suggests that tobacco users may be at increased risk of developing severe COVID-19 symptoms and may be less likely to mount a robust immune response to vaccinations [4] [5] [6] . However, pandemic-related restrictions have caused a reduction in the use of non-urgent clinical care 7 , as well as a shift to virtual treatment to reduce unnecessary contacts between patients and providers. The effects of these changes on use of tobacco cessation treatment are largely unknown. Pre-pandemic efforts to offer virtual care were generally concerned with improving cost-effectiveness or with reaching people with limited access to treatment, such as those in remote areas. Telemedicine interventions for substance use disorders have shown high patient satisfaction, appear to be effective, and may have better retention rates due to reduced travel time 8 . Remote care may be particularly beneficial to rural participants, who otherwise must often travel considerable distances to receive inperson care 9 . Although virtual care offers many advantages, it also has the potential to erect new barriers. Its effectiveness may vary with patients' age, culture, literacy, educational attainment, and socioeconomic status 7, 10 , and accessing care remotely may require skills or access to technology that some patients may lack. This is of special interest in the case of tobacco cessation services, whose patients often have notably lower income and education than others, as well as higher prevalence of mental, physical, and substance-related health problems 11 . On March 17, 2020, Ontario declared a state of emergency due to the COVID-19 pandemic and closed many non-essential services. This led to a rapid virtualization of services in primary care A c c e p t e d M a n u s c r i p t throughout the province. The purpose of this analysis was to examine the effects of these changes on service use for existing and subsequent enrolments in a large smoking-cessation program. We have previously analyzed COVID-related changes in the number of enrolments and clinical visits and on outcomes. This work has shown that total enrolments fell sharply at the beginning of the pandemic; that patients enrolling during the pandemic were slightly more likely to report health problems 12 ; and that people exposed to treatment during this period (including those enrolling before the pandemic) had a slightly lower probability of quitting smoking successfully 13 . In this report, we explore the use of services during the pandemic: whether levels of engagement and total care received per person rose or fell, and whether the associations of treatment use with patient characteristics changed. Here, we are particularly concerned with issues of accessibility related to health or socioeconomic status. We drew data from the Smoking Treatment for Ontario Patients (STOP) program, a large smoking cessation program in Ontario, Canada which, before the COVID-19 pandemic, enrolled about 20,000 smokers per year. STOP provides counseling and up to 26 weeks of Nicotine Replacement Therapy (NRT) at no cost to the participant. Before the pandemic, enrollees were only treated in-person at participating clinics, which included family health teams (physician-led primary care practices), addiction agencies, community health centers, and nurse practitioner-led clinics. In this analysis, we excluded addictions agencies, which offer a different model of care than other clinics, including (in some cases) residential treatment. The population served by STOP is, on average, older than the general Ontario population, and has lower levels of income and education (Table 1) 14 . The exact operational responses to the pandemic among STOP varied across providers and over time, and are not known in detail. In general, however, Ontario primary care providers shifted rapidly to remote treatment following the state of emergency 7, 15 . An informal survey of STOP partners in June A c c e p t e d M a n u s c r i p t 2020 also indicated that they had reduced in-person visits, increased telehealth options, and begun providing NRT via mail or distanced pickup. For this analysis, we included participants who enrolled between April 11 th 2016 (when several important variables were added to the STOP baseline survey) and December 3, 2020. There were 82,389 participants over the study period. We excluded 972 participants who enrolled but had no recorded clinical contacts and 1358 who did not provide informed consent. STOP participants are also able to re-enroll in the program one year after their previous enrolment. We identified repeat enrolments through probabilistic deduplication and kept only the most recent enrolment for each individual. This meant removing a further 14,494 enrolments, leaving an analysis sample of 65,565. We used date of enrolment to define three time periods: more than 6 months before the pandemic; from 6 months before to the onset of the pandemic itself (i.e., people enrolling before the pandemic but potentially using treatment during it); and during the pandemic period. Enrolments during the first period serve as historical controls, while those in the second would have received varying amounts of their treatment after the pandemic began and the third would have received all of their care after pandemic-related changes were in place. Our primary outcome in this analysis was the total number of weeks of NRT received. We also examined the number of clinical visits attended and the total minutes of counseling received (summed across all visits). In all cases, we consider treatment visits within 6 months of enrolment. We evaluate service use primarily in terms of weeks NRT received because the types and dosages of NRT provided vary with patients' preferences and severity of tobacco dependence, while the number of visits (and the amount of NRT supplied at each visit) will vary with the provider and, particularly in the pre-COVID period, with the physical accessibility of treatment. At each clinical visit, providers A c c e p t e d M a n u s c r i p t are asked to record the number of weeks the provided NRT is intended to last, as well the as amount of counseling provided (in minutes). These values were recorded in the same way throughout the study period. We summed these totals for the 6 months of study coverage to produce our outcomes for the total amount of treatment received. In our exploratory analysis of changes in the associations between treatment use and patient characteristics, we chose variables a priori, based on existing studies and previous research with the STOP dataset. We included age, gender, distance to treatment, neighbourhood marginalization, age at smoking initiation, body mass index, self-rated importance of quitting and confidence in ability to quit (both rated 1=lowest to 10=highest), diagnosed substance use disorder, diagnosed physical health condition (heart disease, stroke, cancer, diabetes, or chronic obstructive pulmonary disease), diagnosed mental health condition (schizophrenia, bipolar disorder, depression, anxiety), hazardous drinking (AUDIT-C score), current depressive symptoms (PHQ-2 score), cigarettes smoked per day, household income (8 ordered categories), completion of secondary school, being permanently unable to work, employment status, time to first cigarette after waking (under 30 minutes v. 30 minutes or more), past quit attempts (>5 v. 5 or fewer), rural status (>=40 on the rurality index of Ontario), past-month use of cannabis, past-month use of opioids, ability to handle demands and ability to handle unexpected problems (both rated 1="poor" to 5="excellent"), and Indigenous status (First Nations, Metis, or Inuit). To test whether associations between care use and patient characteristics changed when the pandemic began, we interacted this term with each of the patient variables listed above. We then calculated the joint significance of these interaction terms (as we used multiple imputation, we were unable to conduct a likelihood ratio test for nested models). If the shift to remote care added or removed A c c e p t e d M a n u s c r i p t important barriers that are specific to particular groups of people, then we would expect the effects of these interactions, considered together, to be statistically significant. Finally, we explored specific interactions by adding them individually to the main effects model. Due to the large number of tests performed, we evaluated the significance of these terms using the stepdown method of Holm 16 . In this approach, the lowest p-value is initially compared to a significance threshold of 0.05/M, where M is the number of tests conducted. In our case, this value is 0.0019. The threshold then rises to 0.05/(M-1) and the second-lowest p-value is compared to this new value. This process continues until a non-significant effect is reached. We used Stata 16 and Python 3.8 in our analyses. Patient characteristics are sporadically missing due to non-response on the baseline survey (Table 1) . Proportions missing were highest for our two stress items (ability to handle demands and ability to handle daily problems) and for household income. In these cases, missingness is likely to be largely completely-at-random: the stress items were added to the baseline survey in 2018, and are missing before this time; and missingness for income is very highly concentrated within specific sites and practitioners, which indicates that some providers simply do not ask it. To address missing data, we performed random-forest multiple imputation (RFMI) using the miceforest Python package 17 . RFMI has been shown to perform well, and is more likely than other approaches to preserve complex patterns, including interactions, that may exist in the original data 17 . We generated 20 imputed datasets. A c c e p t e d M a n u s c r i p t Table 1 . Models of change over time (Figure 1) showed that mean weeks of NRT did not change markedly for pre-pandemic enrolments, but rose by about 1 week for enrolments immediately after the state of emergency declaration. After this time, NRT provision declined, but remained significantly above its pre-pandemic level (end of study: predicted mean = 9.4, 95% CI = 9.0, 9.8; September 17, 2019: predicted mean = 8.9, 95% CI = 8.7, 9.1; difference = 0.52, 95% CI = 0.19, 0.85, p=0.002). The mean number of clinical contacts per treatment episode fell slightly for people who enrolled before the pandemic, but recovered for those enrolling after the state of emergency was declared. There was no significant difference between the pre-pandemic level and study end (difference = 0.091; 95% CI = -0.03, 0.21; p=0.13). Counseling minutes fell for pre-pandemic enrolments, and remained lower until the study end (difference = -8.9; 95% CI = -11.3, =6.5; p<0.001). Our omnibus test of interaction terms produced a marginally significant (given the large sample size) p-value: F(25,14864.4)=1.66, p=0.02. We therefore explored individual interactions for importance by adding them individually to the baseline model. Four interactions met our criterion for significance: those for income, area marginalization, inability to work, and the PHQ-2 ( Table 2) . Before the pandemic, more NRT was received by people with higher incomes, lower levels of area marginalization, and lower levels of depression, while there was no association for the PHQ-2. During the pandemic, these effects were smaller or were reversed, with more treatment received by people unable to work and no significant variation for the other three variables. Among primary care clinics participating in a large smoking-cessation program, the rapid change to remote treatment was associated with a reduction in the amount of counseling provided, at least a temporary increase in the weeks of NRT supplied to patients, and no change in the number of A c c e p t e d M a n u s c r i p t clinical contacts per person. Associations with patient characteristics suggest that the effects of some barriers to treatment use may have been reduced during the pandemic. The temporary increase in NRT supplied after the state of emergency (without a corresponding increase in the number of visits) implies that providers initially responded by supplying more NRT at each visit, perhaps due to uncertainty about the logistics of future NRT provision. Similar increases in prescription rates have been seen throughout the pandemic in Ontario nursing homes 18, 19 . With time, however, clinics began to revert to their usual pattern of dosing, and were also able to maintain their existing level of support in terms of clinical contacts. The one area in which care changed notably was in the amount of counseling delivered, which fell by about 14% and has not since recovered. Although the provision of counseling by phone or video conferencing has been extensively studied and has generally been shown to be practical and effective 20 , the abrupt transition to this form of care by providers and patients accustomed to, or expecting, in-person interaction may have created difficulties 21,22 . These may be related to the lack of immediacy in remote counseling, or to the simple physical discomfort of long phone calls or other remote conversations. It is also conceivable, however, that additional support was sometimes supplied during the pandemic without being recorded in the STOP data collection platform. Our results on associations between patient characteristics and use of care implies that the effects of barriers to care were not increased, and may actually have been lessened, by the transition to remote treatment. Accessing care remotely can reduce barriers and difficulties associated with caregiving responsibilities, missed work, stigma, or transport 23, 24 . These barriers are particularly salient for the groups identified in our analysis, who may have poor transportation options and a limited ability to forgo wages, and who may also need to attend clinical appointments for other conditions. We also found no evidence that the transition to remote care had net negative effects on service use for any identifiable groups in our sample. Although we lack detailed data on the response of individual clinics A c c e p t e d M a n u s c r i p t to the pandemic, providers have reported in informal discussions that they provided care largely by telephone; and, given the abrupt change to remote treatment, most would not have been immediately equipped to offer video calls or other options. This may be significant, as telephone treatment does not present the same technological barriers as digital care, which often require computer skills and access to computer hardware and high-speed internet. Telephone care, moreover, has generally been shown to be effective for smoking cessation 25 . It is also possible that providing multiple modalities of care offers a further advantage: for clinics that remained open, people who were comfortable with remote care may willingly have selected this option, which may have preserved in-person treatment for those who preferred it. In other work, we have shown that the proportion of patients who quit smoking successfully declined somewhat during the pandemic, and remained below pre-pandemic levels for enrolments down at least to July, 2020 13 . As both the treatment available and the wider context changed simultaneously, however, we were unable to conclude which was responsible, and fully understanding the implications of remote care for smoking cessation will probably take considerable time. Our present findings, however, strongly imply thatdespite widespread concerns about the erection of new barriers to treatment during the pandemicthere were no substantial negative effects of remote care on treatment equity in our context. The primary health care system in Ontario was remarkably resilient in this respect, with providers adapting rapidly to the demands of health care during the pandemic. The apparent decrease in counseling provided is a possible cause for concern, however, and is a potential focus for quality improvement efforts. We lack data on some patient-level variables that may be associated with use of care, including race and ethnicity (with the exception of Indigenous status), marital status, and immigrant status. Our data on counseling received also reflects the total amount recorded by practitioners during clinical contacts A c c e p t e d M a n u s c r i p t recorded in the data collection system; as we have noted, it is possible that some additional support was not recorded. In our context, and in others, the COVID-19 pandemic changed the treatment model for smoking cessation dramatically and abruptly. Programs appear to have reduced the amount of counseling provided, but did not reduce their usual levels of care with respect to number of clinical contacts and amount of NRT supplied. The changes they have made appear, however, not to have created new inequities, and may have helped to reduce existing disparities. A c c e p t e d M a n u s c r i p t Figure 1 . Predicted means of NRT weeks, counseling minutes, and clinical visits over the first 6m of treatment, by date of enrolment, adjusted for clinic and seasonality. M a n u s c r i p t M a n u s c r i p t 5%) 3,125 (5%) Currently employed 8%) 2,060 (3%) First nations, Metis Distance to clinic (km) 10.8 (15.2) 10 Smoking cessation during COVID-19: the top to-do list A Call to Action: Commercial Tobacco Smoking Cessation Support as a Priority for Health Care Services During the COVID-19 Pandemic Protecting vulnerable groups from tobacco-related harm during and following the COVID-19 pandemic. Health Promot Chronic Dis Prev Can The effect of smoking on COVID-19 severity: A systematic review and meta-analysis COVID-19 and smoking: A systematic review of the evidence Age and smoking predict antibody titres at 3 months after the second dose of the BNT162b2 COVID-19 vaccine Shifts in office and virtual primary care during the early COVID-19 pandemic in Ontario Telehealth-delivered group smoking cessation for rural and urban participants: Feasibility and cessation rates Using technology in the delivery of mental health and substance abuse treatment in rural communities: a review Virtual care: enhancing access or harming care Perceived barriers to smoking cessation in selected vulnerable groups: a systematic review of the qualitative and quantitative literature Changes in the reach of a smoking cessation program in Ontario, Canada, during the COVID-19 pandemic: a cross-sectional study Effect of COVID-19 on smoking cessation outcomes in a large primary care treatment programme: an observational study Variability in patient sociodemographics, clinical characteristics, and healthcare service utilization among 107,302 treatment seeking smokers in Ontario: A cross-sectional comparison Changes in family medicine visits across sociodemographic groups after the onset of the COVID-19 pandemic in Ontario: a retrospective cohort study A simple sequentially rejective multiple test procedure Comparison of random forest and parametric imputation models for imputing missing data using MICE: a CALIBER study Comparison of Medication Prescribing Before and After the COVID-19 Pandemic Among Nursing Home Residents in Ontario, Canada Assessment of psychotropic drug prescribing among nursing home residents in Ontario, Canada, during the COVID-19 pandemic Nicotine replacement therapy versus control for smoking cessation Virtually perfect? Telemedicine for COVID-19 Use of telehealth during the COVID-19 pandemic: scoping review Substance Use Disorder Treatment via Telemedicine During Coronavirus Disease Treatment for anxiety and depression via clinical videoconferencing: evidence base and barriers to expanded access in practice Telephone counselling for smoking cessation 1 Based on marginal fixed effects. Differences shown:Income: weeks for "no reported income" minus weeks for "$100,000 or more" Marginalization: weeks for lowest area marginalization quintile minus highest quintile Inability to work: weeks if "yes" minus weeks if "no" PHQ-2: weeks for score of 7 minus weeks for score of 1. A c c e p t e d M a n u s c r i p t Household income <$40,000 3 23,602 (58.3%) 25,088 (38%) Ability to handle demands (1-5) 4 3.4 (1.0) 30,064 (46%)Ability to handle problems (1-5) 4 3.1 (1.1) 30,233 (46%) 1 5 ordered categories, with 1=low marginalization and 5=high marginalization. 2 10 ordered categories, with 1=lowest and 10=highest. 3 Collected as 8 ordered categories: no reported income, <$10k, 10k-20k, 20k-40k, 40k-60k, 60k- 80k, 80k-100k; over 100k.