Long-term impact of the COVID-19 pandemic on self-management of chronic conditions among high-risk adults in the USA: protocol for the C3 observational cohort study

Abstract

Introduction

COVID-19 is an unprecedented public health threat in modern times, especially for older adults or those with chronic illness. Beyond the threat of infection, the pandemic may also have longer-term impacts on mental and physical health. The COVID-19 & Chronic Conditions (‘C3’) study offers a unique opportunity to assess psychosocial and health/healthcare trajectories over 5 years among a diverse cohort of adults with comorbidities well-characterised from before the pandemic, at its onset, through multiple surges, vaccine rollouts and through the gradual easing of restrictions as society slowly returns to ‘normal’.

Methods and analysis

The C3 study is an extension of an ongoing longitudinal cohort study of ‘high-risk’ adults (aged 23–88 at baseline) with one or more chronic medical conditions during the COVID-19 pandemic. Five active studies with uniform data collection prior to COVID-19 were leveraged to establish the C3 cohort; 673 adults in Chicago were interviewed during the first week of the outbreak. The C3 cohort has since expanded to include 1044 participants across eight survey waves (T₁–T₈). Four additional survey waves (T₉–T₁₂) will be conducted via telephone interviews spaced 1 year apart and supplemented by electronic health record and pharmacy fill data, for a total of 5 years of data post pandemic onset. Measurement will include COVID-19-related attitudes/behaviours, mental health, social behaviour, lifestyle/health behaviours, healthcare use, chronic disease self-management and health outcomes. Mental health trajectories and associations with health behaviours/outcomes will be examined in a series of latent group and mixed effects modelling, while also examining mediating and moderating factors.

Ethics and dissemination

This study was approved by Northwestern University’s Feinberg School of Medicine Institutional Review Board (STU00215360). Results will be published in international peer-reviewed journals and summaries will be provided to the funders of the study.

Correspondence to Dr Rebecca Lovett; [email protected]

Strengths and limitations of this study

• All C3 participants have ‘pre-COVID’ baseline data on a number of psychosocial factors, lifestyle and health behaviours, healthcare use, patient-reported and chronic disease outcomes. This allows us to examine patient trajectories longitudinally from before, during and after the pandemic.

• The C3 study includes data from multiple sources, including phone surveys using validated assessments but also electronic health records (including patient portal use) and pharmacy fill data. This multimethod approach enables comprehensive measurement using both patient-reported and objective measures.

• In the context of COVID-19, the C3 study measures many understudied, potentially modifiable patient factors (eg, health literacy, health activation, tangible support needs, routine, technology access, etc) in addition to mental and physical health.

• The primary limitation will be loss to follow-up and missing data points that may introduce attrition bias.

Introduction

The SARS-CoV-2 and resultant COVID-19 first emerged in China in 2019 and rapidly spread worldwide. The COVID-19 pandemic has since become the largest public health threat of modern times. As of June 2023, over 750 million confirmed cases of COVID-19 and nearly 7 million deaths have been recorded globally.¹ Beyond the direct consequences to personal health associated with acquiring COVID-19, the pandemic has also caused disruptions among nearly all facets of daily life since its onset, with state mandates closing or limiting businesses, requiring individuals to social distance and/or wear masks and changing or limiting access to in-person healthcare and other support services.^2–4 These restrictions have been partially lifted and then reinstated over time as new surges of the virus have emerged, contributing to a sense of uncertainty. Although the availability of vaccines and increasing population immunity has tempered COVID-19 risk over time, allowing a slow return to ‘normal’ pre-pandemic life, concerns remain surrounding the extended, pandemic ‘spillover effects’ on individuals’ long-term physical and mental health.⁵ Retrospective investigations among previous public health crises, such as the 1918 ‘Spanish flu’ pandemic and China’s Severe Acute Respiratory Syndrome and Influenza A virus subtype H7N9 (‘avian flu’) outbreaks found sustained fear and anxiety in the community 5 years later, along with longstanding changes in societal interactions and health behaviours.^6–10

Such concerns have been particularly salient for older adults and those with pre-existing chronic conditions, who at the onset of the COVID-19 pandemic were quickly identified to be at higher risk of acquiring the virus and experiencing adverse consequences if infected, and therefore important targets of disease mitigation efforts.^{11 12} Initial concerns surrounded increased loneliness and isolation due to lockdown measures, as well as ongoing community disruptions associated with other precautionary measures.¹³ Additionally, these patients often contend with considerable, persistent disease and treatment burdens, of which existing self-management demands may have been further complicated by the pandemic. A sizeable body of literature exists documented that, even prior to COVID-19, adults with comorbidities already had difficulty navigating health systems, engaging healthcare providers, managing and adhering to prescribed regimens and assuming daily oversight in self-monitoring health, resulting in greater risks for poorer physical and mental health.^14–31

A recent systematic review and meta-analysis of 65 longitudinal cohort studies found that mental health symptoms among individuals with comorbidities increased from pre-pandemic levels during the first year of the pandemic, while death and infection rates remained high, and many mitigation measures were still in place.³² Additional cross-sectional investigations have also noted limited healthcare access, use or avoidance, as well as uptake of unhealthy lifestyle behaviours and inadequate self-management of chronic medical conditions among this population during this time.^33–35 Now, at a time of easing risk and restrictions, concerns have been raised for persistent ‘COVID-19 Anxiety Syndrome’, with many adults, especially if immunocompromised, having difficulty with societal re-entry.³⁶ However, the prolonged implications to mental health due to the pandemic and any longstanding, residual effects on social behaviour, lifestyle, healthcare use, self-management behaviours and health outcomes are still unclear.

A conceptual framework

Stress, anxiety and depression all are known, significant determinants of physical health.^37–39 Cohen et al summarise in their Stage Model of Stress and Disease the epidemiological, psychological and biological mechanisms linking stressful life experiences to both increased risk of developing a new physical condition and also worsening of existing chronic conditions (figure 1).⁴⁰ Exposure to environments and events that are perceived as threatening may generate a negative emotional reaction, which in turn activate biologic responses, impacting decision-making and behaviour. Both emotional (ie, depression, anxiety) and biological triggers of stress are linked to detrimental health behaviours such as increased smoking, alcohol consumption and substance use, poor sleep, sedentary activity, weight gain and limited adherence to treatment. Self-management support interventions promoting patient health literacy, motivation and self-efficacy have been effective in reducing stress and improving chronic disease outcomes.^41–43

Figure 1. Stage model of stress and disease.

The pandemic should be viewed similar to a disaster or large-scale catastrophic life event: it has threatened harm and death to a large group of people, affected social processes, disrupted services and social networks and has had both direct and secondary consequences in terms of mental and physical health.^{44 45} Studies of prior disasters demonstrate that while the majority of exposed individuals demonstrate resilience, many experience psychological distress over time and slowly recover. Yet some have sustained, clinically relevant, psychological impairment.^46–48 This can range from post-traumatic stress, major depression, substance use and generalised anxiety disorders, among others. According to Goldmann and Galea (figure 2), whether one demonstrates resilience during a traumatic event like COVID-19, experiences stress yet recovers with time, or has clinically relevant problems is determined by: (1) ‘pre-disaster’ or ‘pre-trauma’ risk/protective factors, (2) exposures and experiences during the event (‘peri-disaster’ or ‘peri-trauma’) and (3) residual life stressors and social support available after the pandemic (‘post-disaster’ or ‘post-trauma’).⁴⁸ Pre-trauma risk factors leading to poor mental health outcomes include female sex, low socioeconomic status (SES), prior traumatic experiences, poorer physical or mental health and inadequate social support. Peri-trauma risk factors reflect the nature of the exposure (eg, COVID-19 infection, or experiencing grief/loss from a COVID-19 death). Change in SES circumstances or loss of social relationships are also meaningful losses. Finally, ‘post-trauma’ risk factors include any sustained changes to SES, social interactions or support.

Figure 2. Pre-trauma, peri-trauma and post-trauma framework.

Study aims and hypotheses

There are many limitations in disaster research that hinder understanding who will have longer-term, mental health consequences from COVID-19. The COVID-19 and Chronic Conditions (‘C3’) study is specifically focused on adults at high COVID-19 risk due to either age or underlying medical conditions, and is novel in having a ‘pre-COVID’ (pre-disaster) baseline, along with measurement beginning at the initial week of the outbreak and throughout the pandemic. Additional follow-up during the post-COVID-19 outbreak period will be a unique contribution given the majority of research utilises ‘post-disaster’-only designs.⁴⁹ Given known associations between mental and physical health outcomes, the C3 cohort study will offer additional understanding of the extended mental health effects of the pandemic, in addition to novel investigations of COVID-19’s ‘downstream’ impact on social behaviour, lifestyle, healthcare use, self-management behaviours and health outcomes.

The study aims and hypotheses are listed in table 1.

Table 1

C3 aims and hypotheses

Our primary aims and hypotheses (H) are to:
Aim 1	Evaluate changes in lifestyle, health behaviours, healthcare use, health status and chronic disease outcomes from a pre-pandemic baseline through 5 years after the onset of COVID-19.
H1	The prevalence of harmful alcohol/substance use, inadequate treatment adherence, underuse of routine & preventive health services, poor health status and uncontrolled chronic conditions will be higher during the pandemic compared with a ‘pre-COVID-19’ baseline.
Aim 2	Determine the extent to which stress, anxiety and depression contribute to poor health status and chronic disease outcomes through 5 years after the pandemic’s onset.
H2	Intermittent or persistently higher levels of stress, anxiety and depression during/after the pandemic will be associated with poor health status and uncontrolled chronic conditions.
Our secondary aim is to:
Aim 3	Identify factors that mediate or moderate associations between stress, anxiety and depression during/after the pandemic with health status and chronic disease outcomes.
H3	Harmful alcohol/substance use, physical inactivity, weight gain, sleep disturbance, inadequate adherence and underuse of routine & preventive health services during/after the pandemic will mediate associations.
H4	Greater health literacy, chronic disease self-efficacy, health activation and social support prior to, during and/or after the pandemic will moderate associations.
Our exploratory aim is to:
Aim 4	Explore whether health disparities by age, sex, race, ethnicity or socioeconomic status emerge or worsen during/after the pandemic and the contributing role of stress, anxiety and depression.

Methods and analysis

Study design

On 13 March 2020, our team at Northwestern University launched the C3 study, leveraging five active NIH projects (table 2). The five ‘parent studies’ had uniform data collection on a range of patient-reported outcomes within 1 year prior to COVID-19, as well as permission for access to patient electronic health records (EHRs) and pharmacy records to capture prescription fill data. The initial C3 study has been described previously.⁵⁰

Table 2

C3 parent studies

Parent study (NIH Project Number)	Design	Sample characteristics
		Language	Clinical inclusion criteria	Clinical exclusion criteria	Setting
Health Literacy and Cognitive Function Among Older Adults (R01AG030611)	Cohort	English	Aged 55–75 at study onset	Severe, uncorrectable vision, hearing or cognitive impairments Limited English proficiency	1 academic internal medicine clinic, 5 FQHCs
A Universal Medication Schedule to Promote Adherence to Complex Drug Regimens (R01AG046352)	Clinical Trial	English or Spanish	Aged ≥50 at study onset Taking ≥5 long-term Rx medications	Severe, uncorrectable vision, hearing or cognitive impairments	1 academic internal medicine clinic, 1 FQHC
Transplant Regimen Adherence for Kidney Recipients by Engaging Information Technologies: The TAKE IT Trial (R01DK110172)	Clinical Trial	English	Aged ≥21 at study onset 5 weeks to 24 months post-kidney transplant Taking ≥3 long-term Rx med	Severe, uncorrectable vision, hearing or cognitive impairments Limited English proficiency	1 academic transplant centre
EHR-Based Universal Medication Schedule to Improve Adherence to Complex Regimens (R01NR015444)	Clinical Trial	English	Aged ≥21 at study onset T2DM Haemoglobin A1C value ≥7.5% at study onset Taking ≥5 long-term Rx med	Severe, uncorrectable vision, hearing or cognitive impairments Limited English proficiency	5 academic internal medicine clinics
Self-Management Behaviours among COPD Patients with Multimorbidity (R01HL126508)	Cohort	English	Aged ≥40 at study onset COPD and/or HTN, T2DM	Severe, uncorrectable vision, hearing, or cognitive impairments Limited English proficiency	2 academic internal medicine clinics

COPD, chronic obstructive pulmonary disease; EHR, electronic health record; FQHC, Federally Qualified Health Centre; HTN, hypertension; Rx, prescription; T2DM, type 2 diabetes mellitus.

During the first week of the outbreak, 673 adults were recruited and interviewed via phone in Chicago, Illinois, USA (T₁: 13–20 March 2020). A second wave of interviews was conducted 2 weeks later (T₂: 27 March–3 April 2020), followed by a third wave soon after (T₃: 27 April–3 May 2020), which added 199 patients from Mount Sinai Hospital practices in New York City (NYC) linked to two of the parent studies (n=872). A National Institute on Aging COVID-19 supplement (R01AG030611-13S1) was awarded in August 2020 to expand the cohort via the parent studies (n=1044) and continue data collection through May 2022 (T₅–T₈). To date, eight survey waves have been completed, with high levels of retention achieved (83%–94%). An additional National Institute on Aging award (R01AG075043) was received in June 2022 to further extend data collection. These present C3 study activities will expand follow-up of the C3 cohort for an additional four waves (T₉–T₁₂) to capture 5 years of post-COVID-19 data capture (figure 3).

Figure 3. C3 study timepoints.

Performance sites

C3 participants were originally recruited from parent studies with multiple academic and community health centre performance sites in Chicago, Illinois, USA: (1) Northwestern Memorial Healthcare (NMHC), consisting of multiple, large academic practices; (2) Access Community Health Network, a Public Health Service 330-funded network of Federally Qualified Health Centers (FQHCs) and (3) Erie Family Health Center, a network of FQHCs comprised of 12 large health centres affiliated with the AllianceChicago EHR user community. The C3 cohort was expanded for T₃ and T₅ with participants from the (4) Mount Sinai Icahn School of Medicine academic practice in NYC.

Subject eligibility and recruitment

Specific eligibility criteria for the parent studies varied, however, all participants were eligible for the initial C3 survey if they had reported at least one chronic medical condition during their parent study interview and had completed their most recent parent study interview in 2019 or later. The inclusion criteria for participation in the present study include having been enrolled in and completed at least one prior C3 survey wave.

The proposed study will mirror the methodology used in previous waves of data collection for the C3 cohort.⁵⁰ Specifically, trained research coordinators (RCs) will conduct telephone-based interviews to collect new data from C3 participants every year, over the course of 4 years (T9–T12). RCs will contact all active C3 participants to invite them to participate in the four additional interviews. RCs will explain the study procedures and will obtain verbal informed consent using procedures approved by the Northwestern University Feinberg School of Medicine Institutional Review Board. Study dates are from May 2022 to May 2027, with data collection for T₉ beginning in Spring 2023.

Data sources

Data will be gathered and merged from the following sources:

Patient interviews

RCs will conduct interviews by telephone using REDCap survey software. Patients will be compensated a total of $95 for completing the four interviews ($20 for T₉ and T₁₀, $25 for T₁₁ and $30 for T₁₂).

EHR data

Clinical data will be requested from each site. International Classification of Diseases, Tenth Edition (ICD-10) and Current Procedural Terminology (CPT) codes will be used to confirm chronic conditions using relevant date ranges.

Patient measurement

Table 3 summarises included measures.

Table 3

C3 study measures and outcomes

Variable	Instrument(s) or measure(s)	Pre-COVID-19	Initial C3 cohort (T1–T8)	R01 C3 cohort (T9–T12)
Sociodemographic and psychosocial characteristics
Sociodemographic	Age, sex, race/ethnicity, education, employment, household income, household composition, marital status	X	X	X
Patient activation	Consumer Health Activation Index (CHAI)	X	X	X
Health literacy	Newest Vital Sign (NVS)	X
Social support	2-item measure of tangible social support; Martin & Park environmental demands	X	X	X
Self-efficacy	Self-Efficacy for Managing Chronic Disease 6-item scale		X	X
COVID-19-related beliefs and actions
Awareness and concern	Perceptions of the virus; feelings about the pandemic		X	X
Preparedness	Personal preparedness; confidence in state/federal response		X	X
Actions	Daily routine; change in plans; obtaining prescriptions, leaving home		X	X
Testing & diagnosis	Ability to obtain COVID-19 test, outcome, diagnosis		X	X
Information-seeking and sources	Amount time/day getting COVID-19 news; information source		X	X
Behavioural factors
Medication adherence	Ask-12; Proportion of Days Covered (pharmacy records)	X	X	X
Harmful alcohol use	Alcohol Use Disorders Identification Test (AUDIT-C)	X	X	X
Cigarette smoking	Behavioural Risk Factor Surveillance System (BRFSS) Cigarette Smoking items	X	X	X
Nutrition	Rapid Eating Estimate for Participants (REAPS)		X	X
Use of health information technology and Health Services
Portal use	Use of patient portal; Type of patient portal utilisation	X	X	X
Health information seeking	Health Information National Trends Survey (HINTS) and Pew Research Survey items on technology use, information seeking	X	X	X
Telehealth experiences	Telehealth visit (yes/no); satisfaction and preferences for future care		X	X
Routine and preventive services use	Use of routine and specialty clinic visit, cancer screening (self-report and EHR)	X	X	X
Urgent care	Emergency Department/Urgent Care visits or hospitalizations (self-report and EHR)	X	X	X
Vaccination	Vaccinated (yes/no) for influenza, shingles, COVID-19 in past 12 months (self-report and EHR)	X
Health services use	Routine, preventive and urgent healthcare utilizations (self-report and EHR)		X	X
Physical and mental health
Stress	Cohen 10-item Perceived Stress Scale (PSS-10); PTSD Checklist (PCL-5)		X	X
Isolation and loneliness	UCLA 3-item loneliness scale; 1-item loneliness due to COVID-19		X	X
Sleep health	Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance Short Form		X	X
Cognitive complaints	Everyday Cognition (ECog) subscale		X	X
Anxiety	PROMIS 8-item Anxiety Short Form	X	X	X
Depression	PROMIS 8-item Depression Short Form	X	X	X
Overall mental well-being	WHO Five Well-being Index (WHO-5)		X	X
Health status	PROMIS 10-item Physical Function Short Form; Self-reported overall health; Number of chronic conditions	X	X	X
Regimen complexity	Medication Regimen Complexity Index (MRCI)	X	X	X
Chronic disease outcomes	Haemoglobin A1c, blood pressure, cholesterol and others as relevant (EHR)	X	X	X

EHR, Electronic Health Record; PTSD, Post-Traumatic Stress Disorder; UCLA, University of California Los Angeles.

Sociodemographic and psychosocial characteristics

All parent studies collected self-report of patient demographics (age, sex, race, ethnicity) and socioeconomic position (household income, employment status, education). Socioeconomic status variables are also routinely collected during C3 waves to note any changes. Additionally, other measures will include:

Patient activation

The Consumer Health Activation Index⁵¹ was administered during the parent studies to assess patients’ degree of ‘activation’ or motivation to participate in healthcare decisions and actions.

Health literacy

Parent studies assessed health literacy using the Newest Vital Sign⁵², a reliable screening tool used to determine risk for limited health literacy.

Social support

Perceived adequacy of tangible social support was assessed using a 2-item social support scale as part of the parent studies; it is also a part of the C3 battery.⁵³

Self-efficacy

The Self-Efficacy for Managing Chronic Disease 6-item Scale examines multiple domains of chronic disease self-management, including symptom control, role function, emotional functioning and communicating with physicians and confidence in completing each activity.⁵⁴

COVID-19-related beliefs and actions

COVID-19 awareness and concern

Awareness of COVID-19 was assessed in early C3 waves using three items that asked whether participants had heard of the coronavirus. Perceived concern for COVID-19 is evaluated by asking subjects to rate, on a scale of 1–10 how serious of a public health threat they felt the coronavirus is or might become. In addition, participants also rate their level of worry about getting the coronavirus (very worried, somewhat worried, a little worried, not worried at all).

COVID-19 preparedness

Participants are asked to self-report their confidence in local, state and federal government to manage the pandemic (very, somewhat, not very or not at all confident), and perceived preparedness to contend with COVID-19 (very, somewhat, a little or not at all prepared).

COVID-19 actions

Participants are asked to describe any changes to their daily routines or plans as a result of COVID-19. Verbatim responses are documented, and responses independently coded.

COVID-19 testing and diagnosis

Participants are asked if they have, had or think they have/had COVID-19, if they sought consult of a healthcare professional about COVID-19, if they were tested and if yes, the result. Acute and chronic COVID-19 symptoms are also measured, as appropriate.

COVID-19 information seeking and sources

Participants self-report the amount of time per day they are getting news or learning about COVID-19 as well as their sources for information about COVID-19.

Health behaviors

Medication adherence

Medication adherence is assessed using (1) the proportion of days covered (PDC) in studies that collect pharmacy records⁵⁵ and (2) the validated Ask-12 measure.⁵⁶ PDC will be calculated by summing the number of days’ supply obtained by a patient during a given time period and dividing by the number of days for which the medication was prescribed.⁵⁵ Adherence will be treated continuously and dichotomously (yes/no; PDC ≥80%). The Ask-12 measures three domains of medication adherence: forgetfulness, treatment beliefs and behaviour. Scores range from 12 to 60, with higher scores indicating poorer adherence.⁵⁶

Harmful alcohol use

The Alcohol Use Disorders Identification Test (Audit-C) includes three questions about alcohol consumption that can detect excessive drinking and active alcohol abuse.⁵⁷

Cigarette smoking

Behavioural Risk Factor Surveillance System 2013 items are used to measure cigarette smoking.⁵⁸

Nutrition

The Rapid Eating Assessment for Participants is 16 items to assess intake of fat, cholesterol, fibre, sugar and selected food groups.⁵⁹

Use of health information technology and health services

Portal use and health information seeking

Data on whether participants registered and accessed the patient portal is collected via self-report and EHR data. If accessed, information on what the portal was used for is extracted.⁶⁰ General health information-seeking behaviours and sources is collected via items from the Health Information National Trends Survey and Pew Research Center’s survey on technology use to evaluate adults’ access to, use of and trust in various sources of health information.^{61 62}

Telehealth experiences

Self-report of having a telehealth visit, either via video or phone, satisfaction and preferences for future clinical visits are assessed using items developed by Polinski et al.⁶³

Vaccination

Data on vaccine acceptance and uptake for both the COVID-19 vaccine as well as the annual influenza vaccine, shingles and pneumococcal is collected via self-report and EHR data.

Health services use

Routine health services use (clinic visits, medical subspecialty), preventive services utilisation (cancer screening, immunisations) and urgent healthcare utilisations (emergency department visits, unplanned hospitalisations) are collected via EHR data and patient self-report.

Physical and mental health

Stress

The Cohen 10-item Perceived Stress Scale⁶⁴ is the most widely used instrument for measuring the perception of stress. We adapted the scale for participants to contextualise responses to COVID-19. The validated, 4-item Post-Traumatic Stress Disorder (PTSD) Checklist is used to assess PTSD symptoms.⁶⁵

Isolation and loneliness

Isolation is assessed by a single item that asks participants to report the frequency they felt isolated due to COVID-19. We also include the University of California Los Angeles (UCLA) 3-item loneliness scale.⁶⁶

Sleep Health

The Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance Short Form 8a assesses sleep quality, sleep depth and restoration associated with sleep.⁶⁷

Cognitive function

The validated Everyday Cognition scale is a self-reported, subjective measure of cognitive functioning and decision making.^{68 69}

Anxiety

The PROMIS 8-item short-form instrument assesses anxiety symptoms.⁷⁰

Depression

The PROMIS 8-item short-form instrument measures depressive symptoms.⁷⁰

Overall mental well-being

The WHO Five Well-being Index (WHO-5) is used as a brief measure of current mental well-being; it includes five items inquiring about mental health over the prior 2 weeks.^{71 72}

Health status

The PROMIS short-form instrument of physical function and self-reported overall health measure health status.⁷⁰

Regimen complexity

The Medication Regimen Complexity Index quantifies the complexity of medications.⁷³

Chronic disease outcomes

Chronic disease management of common conditions including systolic/diastolic blood pressure, haemoglobin A1c, lipid profile (total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein and triglycerides) and kidney function (estimated glomerular filtration rate (eGFR); mL/min) are used. These values and their dates will be obtained quarterly from the EHR for the entire study period, excluding measures obtained during in-hospital encounters (emergency department visits, acute hospital admissions, hospital in or outpatient procedures).

Analysis plan

Analyses, per aim, are described below. In all analyses, we will include sex as a covariate. With Aims 2–4 analyses, sex differences will be specifically examined to determine the presence of disparities, as well as underlying factors driving any found differences. In addition, we will explore potential interactions between sex, race, ethnicity and age.

Aim 1: evaluate changes in lifestyle, health behaviours, healthcare use, health status and chronic disease outcomes from a pre-pandemic baseline through 5 years after the onset of COVID-19

Bivariate descriptive analyses will be conducted to determine whether patient characteristics (age, gender, race/ethnicity, education, income) associated with the most recent outcome prior to the outbreak are potential confounders. Since the time between visits in the parent studies prior to the outbreak differ, data collected prior to the outbreak (time=0) will be defined as first, second, etc. The amount of time data was collected before the C3 T₁ baseline will be adjusted for in models. To examine the prevalence of routine healthcare use pre-COVID-19 versus post-COVID-19 outbreak, average annualised routine visits will be calculated pre-COVID-19 and post-COVID-19. Chronic disease outcome variables will be similarly defined based on either monthly, quarterly or yearly time intervals depending on how often the outcome was captured. Evaluating changes in health behaviours, healthcare use, health status and chronic disease outcomes collected in the survey or as described previously will be assessed using generalised linear mixed models (GLMMs) accounting for repeated measures on participant study visits prior to the COVID-19 outbreak and at each survey wave. Time will be the main variable of interest and predicted means or mean probabilities over time (standardised as days or months prior to March 2020 (negative times), and days or months post (positive times), will be plotted. We will examine the plots for each outcome to see if there is a notable inflection point at time=0 and how means or probabilities vary across time. All models will adjust for parent study, site, as well as age, sex and other covariates found to be potential confounders in preliminary analyses. We will also include COVID-19 infection status as a time-varying covariate in models, but also examine potential moderating effects. In addition, beyond the primary outcomes stated in our hypotheses, we will also examine in a similar manner the social behaviours related to the pandemic captured at each survey wave (eg, social distancing, masking, shopping, attending groups/events, travel, etc).

Aim 2: determine the extent to which stress, anxiety and depression contribute to prolonged, adverse health outcomes through 5 years after the pandemic’s onset

For Aim 2, we will first identify clusters of individuals following similar progressions in psychological distress (stress, anxiety and depressive symptoms) over 5 years post-pandemic onset and classify into trajectory groups. Associations will be examined between trajectories and health outcomes at 5 years post-onset. To group patients into trajectory groups, we will use the PROC TRAJ procedure in SAS V.9.4 to determine differing trajectories of stress, anxiety and depression.^{74 75} This method estimates discrete mixture models on longitudinal data, in our case assuming a censored normal distribution (censored between minimum and maximum scores). The Bayesian Information Criterion will be used to determine the number of discrete trajectories in the data and the degree of the polynomials of the trajectories.

Participants will then be assigned to a trajectory based on posterior probabilities of belonging to each. Once trajectory groups are identified per measure, preliminary bivariate analyses (eg, t-tests, one way ANOVA, χ² tests) will be conducted to determine whether stress, anxiety and depression trajectories are associated with patient characteristics. To examine associations between trajectory groups and health outcomes, we will employ GLMMs, which allow us to incorporate outcome variables that are continuous, dichotomous and count. All models will incorporate random effects to account for repeated measurements over the study time points, with fixed effects for time and any potential confounders from preliminary analyses, and parent study as a mixed effect.

Aim 3: identify factors that mediate or moderate associations between stress, anxiety and depression throughout the pandemic with health status and chronic disease outcomes

Using methods by Iacobucci, we will examine potential mediators of the relationship between trajectories of stress, anxiety and depression to functional health status and chronic disease outcomes (figure 4).^76–78 First, we will examine associations between trajectories and health behaviours using models similar to those described in Aim 2, but with health behaviours as the outcome. If this model holds for a particular mediator, we will proceed by adding it to respective models described in Aim 2 where significant relationships between psychological distress trajectories and functional and chronic disease health outcomes are found. Trajectory membership will be examined to determine whether the addition of mediators eliminates or decreases the strength of the relationship. Interactions between psychological distress trajectories and health behaviours in predicting health status outcomes to assess for moderation.

Figure 4. Aim 3: mediation and moderation models.

Aim 4: explore whether health disparities by age, sex, race, ethnicity or socioeconomic status emerge or worsen during/after the pandemic and the contributing role of stress, anxiety and depression

Separate GLMM models will be employed for each health outcome, one for time points prior to the outbreak and another for time points during/after COVID-19. All models will incorporate random effects to account for repeated measurements over the study time points, with fixed effects for time and any potential confounders from preliminary analyses and parent study as a mixed effect. Pre-outbreak model estimates will first be examined to determine whether disparities exist. If so, Least Square Means (LSM) will be calculated by characteristic and compared with LSM in post-outbreak models. Mediation and moderation models—analogous to models described in Aim 3—will determine the contributing role of psychological distress on any new disparities during/after the pandemic, or differences in the extend of disparity from pre-COVID-19 to periods during/after. Similar to Aims 2 and 3, additional models can then be explored to investigate potential mediators and moderators of disparities.

Power considerations

Estimated sample sizes per wave ranges from 800 to 1000. With n=1000 (estimated sample size for T9), we have enough information to provide 95% CI for means with an actual distance from mean to a limit of 0.062 SD. Similarly, a sample of 800 will provide information to create 95% CIs with an actual distance from mean to a limit of 0.066 SDs. As most measures have clinically meaningful changes of ~1.0 SD, we will be able to track if these changes are clinically relevant throughout the pandemic, and closely examine trends as the COVID-19 pandemic dissipates and/or ends during the period of study. Note that these analyses will include longitudinal data from each participant and use all available data for participants with at least 2 and up to 12 observations post-onset of the pandemic. Estimates of within participant autocorrelations would be speculative, but we do anticipate this will add more information and lead to smaller actual confidence intervals. While it is difficult to power for an interactive (modifying) effect, assuming a minimum of 20% of the sample has reported a positive test for COVID-19, a subsample based on 200 has enough information to provide 95% CI for means with an actual distance from mean to a limit of 0.14 SD; ample to detect a clinically meaningful change in outcomes within this smaller subset.

Methodological issues

The primary limitation of the C3 study will be loss to follow-up and missing data points that may introduce attrition bias. We will employ strategies to minimise participant attrition, including interview reminders, compensation for participant time and annual newsletters sharing publishing C3 findings. The C3 sample is also a convenience sample of primarily urban, older adults with chronic conditions. Thus, our findings may have limited generalizability to other populations.

Patient and public involvement

Patients or the public were not involved in the design, conduct, reporting or dissemination plans of the C3 study.

Ethics and dissemination

Northwestern University’s Feinberg School of Medicine Institutional Review Board (STU00215360) individually reviewed and approved all previous C3 study waves (T₁–T₈) prior to initiation of each wave’s data collection activities. Northwestern University’s Feinberg School of Medicine IRB (STU00215360) also reviewed and approved the current C3 study protocol reported herein, and all planned study waves (T₉–T₁₂) will undergo further review and approval prior to the start of data collection. Results will be published in international peer-reviewed journals and summaries will be provided to the funders of the study.

Ethics statements

Patient consent for publication

Not applicable.

Footnote

Contributors RL contributed to manuscript conception, design of the work, as well as drafted the manuscript and revised it following author feedback. SF, AR, EY, SW-L and AV contributed to manuscript conception and editing of manuscript drafts. MB, JYB, RO, GW, PZ, MMK, SL, LMC, AF and SCB contributed to manuscript conception, design of the work and editing of manuscript drafts. MW contributed to funding acquisition, manuscript conception, design of the work and editing of manuscript drafts. All authors have approved the submitted version.

Funding This work is supported by grants from the National Institutes of Health (R01AG075043-01). The funding agency played no role in the study design, collection of data, analysis or interpretation of data.

Competing interests SCB reports grants from the NIH, Merck, Pfizer, Gordon and Betty Moore Foundation, Retirement Research Foundation for Aging, Lundbeck, Gilead and Eli Lilly via her institution and personal fees from Sanofi, Pfizer, University of Westminster, Lundbeck, Gilead and Luto UK outside the submitted work. MW reports grants from the NIH, Gordon and Betty Moore Foundation, and Eli Lilly, and personal fees from Pfizer, Sanofi, Luto UK, University of Westminster and Lundbeck outside the submitted work. All other authors have no conflicts of interest to disclose.

Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Provenance and peer review Not commissioned; peer reviewed for ethical and funding approval prior to submission.

References

¹ World Health Organization. WHO Coronavirus (COVID-19) dashboard. n.d. Available: https://covid19.who.int/

² Illinois Department of Public Health. Coronavirus disease 2019 (COVID-19). 2020. Available: https://www.dph.illinois.gov/covid19

³ Petrella S, Johnson S, Pratt G. Gov. J.B. Pritzker issues order requiring residents to ‘stay at home’ starting Saturday. Chicago Tribune; 2020.

⁴ Mahase E. Covid-19: mental health consequences of pandemic need urgent research, paper advises. BMJ 2020: m1515. doi:10.1136/bmj.m1515

⁵ Chen J, McGeorge R. Spillover effects of the COVID-19 pandemic could drive long-term health consequences for non-COVID-19 patients [ Health Affairs Blog ]. 2020. Available: https://www.healthaffairs.org/content/forefront/spillover-effects-covid-19-pandemic-could-drive-long-term-health-consequences-non-covid

⁶ Alfani G, Melegaro A. Pandemie d’Italia. Dalla peste nera all’influenza suina: l’impatto sulle società. Milano: Egea, 2010.

⁷ Alfani G, Murphy TE. Plague and lethal epidemics in the pre-industrial world. J Econ Hist 2017; 77: 314–43. doi:10.1017/S0022050717000092

⁸ Almond D. Is the 1918 influenza pandemic over? Long‐term effects of in utero influenza exposure in the post‐1940 U.S. population. J Polit Econ 2006; 114: 672–712. doi:10.1086/507154

⁹ Helgertz J, Bengtsson T. The long lasting influenza: the impact of fetal stress during the 1918 influenza pandemic on socioeconomic attainment and health in Sweden 1968-2012. Demography 2019; 56: 1389–425. doi:10.1007/s13524-019-00799-x

¹⁰ Myrskylä M, Mehta NK, Chang VW. Early life exposure to the 1918 influenza pandemic and old-age mortality by cause of death. Am J Public Health 2013; 103: e83–90. doi:10.2105/AJPH.2012.301060

¹¹ Yang J, Zheng Y, Gou X, et al. Prevalence of comorbidities and its effects in patients infected with SARS-CoV-2: a systematic review and meta-analysis. Int J Infect Dis 2020; 94: 91–5. doi:10.1016/j.ijid.2020.03.017

¹² Bialek S, CDC COVID-19 Response Team. Centers for disease control and prevention. severe outcomes among patients with coronavirus disease 2019 (COVID-19) - United States, February 12-March 16, 2020. MMWR Morb Mortal Wkly Rep 2020; 69: 343–6. doi:10.15585/mmwr.mm6912e2

¹³ Vahia IV, Jeste DV, Reynolds CF. Older adults and the mental health effects of COVID-19. JAMA 2020; 324: 2253. doi:10.1001/jama.2020.21753

¹⁴ Wolf MS, Curtis LM, Waite K, et al. Helping patients simplify and safely use complex prescription regimens. Arch Intern Med 2011; 171: 300–5. doi:10.1001/archinternmed.2011.39

¹⁵ Davis TC, Wolf MS, Bass PF, et al. Literacy and misunderstanding prescription drug labels. Ann Intern Med 2006; 145: 887. doi:10.7326/0003-4819-145-12-200612190-00144

¹⁶ Institute of Medicine. Health literacy: a prescription to end confusion. Washington DC: National Academy Press, 2004.

¹⁷ Wolf MS, Gazmararian JA, Baker DW. Health literacy and functional health status among older adults. Arch Intern Med 2005; 165: 1946–52. doi:10.1001/archinte.165.17.1946

¹⁸ McCarthy DM, Waite KR, Curtis LM, et al. What did the doctor say? Health literacy and recall of medical instructions. Med Care 2012; 50: 277–82. doi:10.1097/MLR.0b013e318241e8e1

¹⁹ Baker DW, Wolf MS, Feinglass J, et al. Health literacy and mortality among elderly persons. Arch Intern Med 2007; 167: 1503–9. doi:10.1001/archinte.167.14.1503

²⁰ O’Conor R, Wolf MS, Smith SG, et al. Health literacy, cognitive function, proper use, and adherence to inhaled asthma controller medications among older adults with asthma. Chest 2015; 147: 1307–15. doi:10.1378/chest.14-0914

²¹ Goodman ER, Federman AD, Martynenko M, et al. Adherence to multiple medications in older adults with asthma. J Am Geriatr Soc 2016; 64: 1133–5. doi:10.1111/jgs.14124

²² Fortin M, Soubhi H, Hudon C, et al. Multimorbidity’s many challenges. BMJ 2007; 334: 1016–7. doi:10.1136/bmj.39201.463819.2C

²³ Clark NM. The multiple challenges of multiple morbidities. Health Educ Behav 2011; 38: 219–21. doi:10.1177/1090198111410554

²⁴ Benner JS, Glynn RJ, Mogun H, et al. Long-term persistence in use of statin therapy in elderly patients. JAMA 2002; 288: 455–61. doi:10.1001/jama.288.4.455

²⁵ Shay CM, Ning H, Allen NB, et al. Status of cardiovascular health in US adults. Circulation 2012; 125: 45–56. doi:10.1161/CIRCULATIONAHA.111.035733

²⁶ Gellad WF, Grenard JL, Marcum ZA. A systematic review of barriers to medication adherence in the elderly: looking beyond cost and regimen complexity. Am J Geriatr Pharmacother 2011; 9: 11–23. doi:10.1016/j.amjopharm.2011.02.004

²⁷ Federman AD, Wolf MS, Sofianou A, et al. Self-management behaviors in older adults with asthma: associations with health literacy. J Am Geriatr Soc 2014; 62: 872–9. doi:10.1111/jgs.12797

²⁸ Boyd CM, Wolff JL, Giovannetti E, et al. Healthcare task difficulty among older adults with multimorbidity. Med Care 2014; 52 Suppl 3: S118–25. doi:10.1097/MLR.0b013e3182a977da

²⁹ Bayliss EA, Ellis JL, Steiner JF. Barriers to self-management and quality-of-life outcomes in seniors with multimorbidities. Ann Fam Med 2007; 5: 395–402. doi:10.1370/afm.722

³⁰ McDermott MM, Liu K, Greenland P, et al. Functional decline in peripheral arterial disease: associations with the ankle brachial index and leg symptoms. JAMA 2004; 292: 453–61. doi:10.1001/jama.292.4.453

³¹ Lindquist LA, Lindquist LM, Zickuhr L, et al. Unnecessary complexity of home medication regimens among seniors. Patient Educ Couns 2014; 96: 93–7. doi:10.1016/j.pec.2014.03.022

³² Robinson E, Sutin AR, Daly M, et al. A systematic review and meta-analysis of longitudinal cohort studies comparing mental health before versus during the COVID-19 pandemic in 2020. J Affect Disord 2022; 296: 567–76. doi:10.1016/j.jad.2021.09.098

³³ Chen J, McGeorge R. Spillover effects of the COVID-19 pandemic could drive long-term health consequences for non-COVID-19 patients. Bethesda, Maryland Health Affairs; 2020. Available: https://www.healthaffairs.org/do/10.1377/hblog20201020.566558/full

³⁴ Kearney A, Lopes L, Brodie M. Vaccine hesitancy among hispanic adults [ Kaiser Family Foundation ]. 2021. Available: https://www.kff.org/coronavirus-covid-19/poll-finding/vaccine-hesitancy-among-hispanic-adults

³⁵ Kretchy IA, Asiedu-Danso M, Kretchy J-P. Medication management and adherence during the COVID-19 pandemic: perspectives and experiences from low-and middle-income countries. Res Social Adm Pharm 2021; 17: 2023–6. doi:10.1016/j.sapharm.2020.04.007

³⁶ Nikčević AV, Spada MM. The COVID-19 anxiety syndrome scale: development and psychometric properties. Psychiatry Res 2020; 292: 113322. doi:10.1016/j.psychres.2020.113322

³⁷ Thoits PA. Stress and health: major findings and policy implications. J Health Soc Behav 2010; 51 Suppl: S41–53. doi:10.1177/0022146510383499

³⁸ Garfin DR, Thompson RR, Holman EA. Acute stress and subsequent health outcomes: a systematic review. J Psychosom Res 2018; 112: 107–13. doi:10.1016/j.jpsychores.2018.05.017

³⁹ Cohen S, Janicki-Deverts D, Miller GE. Psychological stress and disease. JAMA 2007; 298: 1685. doi:10.1001/jama.298.14.1685

⁴⁰ Cohen S, Gianaros PJ, Manuck SB. A stage model of stress and disease. Perspect Psychol Sci 2016; 11: 456–63. doi:10.1177/1745691616646305

⁴¹ Wang T, Tan J-Y, Xiao LD, et al. Effectiveness of disease-specific self-management education on health outcomes in patients with chronic obstructive pulmonary disease: an updated systematic review and meta-analysis. Patient Educ Couns 2017; 100: 1432–46. doi:10.1016/j.pec.2017.02.026

⁴² Lorig KR, Sobel DS, Stewart AL, et al. Evidence suggesting that a chronic disease self-management program can improve health status while reducing hospitalization: a randomized trial. Med Care 1999; 37: 5–14. doi:10.1097/00005650-199901000-00003

⁴³ Wolf MS, Seligman H, Davis TC, et al. Clinic-based versus outsourced implementation of a diabetes health literacy intervention. J Gen Intern Med 2014; 29: 59–67. doi:10.1007/s11606-013-2582-2

⁴⁴ Lowe SR, Raker EJ, Zacher ML. Extremes in context: a life-course approach to disaster mental health. One Earth 2020; 2: 497–9. doi:10.1016/j.oneear.2020.05.022

⁴⁵ Raker EJ, Zacher M, Lowe SR. Lessons from hurricane Katrina for predicting the indirect health consequences of the COVID-19 pandemic. Proc Natl Acad Sci U S A 2020; 117: 12595–7. doi:10.1073/pnas.2006706117

⁴⁶ Farfel M, DiGrande L, Brackbill R, et al. An overview of 9/11 experiences and respiratory and mental health conditions among world trade center health registry enrollees. J Urban Health 2008; 85: 880–909. doi:10.1007/s11524-008-9317-4

⁴⁷ Norris FH, Tracy M, Galea S. Looking for resilience: understanding the longitudinal trajectories of responses to stress. Soc Sci Med 2009; 68: 2190–8. doi:10.1016/j.socscimed.2009.03.043

⁴⁸ Goldmann E, Galea S. Mental health consequences of disasters. Annu Rev Public Health 2014; 35: 169–83. doi:10.1146/annurev-publhealth-032013-182435

⁴⁹ Galea S, Maxwell AR. Methodological challenges in studying the mental health consequences of disasters. In: Norris FH, Galea S, Neria Y, eds. Mental Health and Disasters. Cambridge: Cambridge University Press, 2009: 579–93.

⁵⁰ Wolf MS, Serper M, Opsasnick L, et al. Awareness, attitudes, and actions related to COVID-19 among adults with chronic conditions at the onset of the U.S. outbreak: a cross-sectional survey. Ann Intern Med 2020; 173: 100–9. doi:10.7326/M20-1239

⁵¹ Wolf MS, Smith SG, Pandit AU, et al. Development and validation of the consumer health activation index. Med Decis Making 2018; 38: 334–43. doi:10.1177/0272989X17753392

⁵² Weiss BD, Mays MZ, Martz W, et al. Quick assessment of literacy in primary care: the newest vital sign. Ann Fam Med 2005; 3: 514–22. doi:10.1370/afm.405

⁵³ Woloshin S, Schwartz LM, Tosteson AN, et al. Perceived adequacy of tangible social support and health outcomes in patients with coronary artery disease. J Gen Intern Med 1997; 12: 613–8. doi:10.1046/j.1525-1497.1997.07121.x

⁵⁴ Lorig KR, Sobel DS, Ritter PL, et al. Effect of a self-management program on patients with chronic disease. Eff Clin Pract 2001; 4: 256–62.

⁵⁵ Choudhry NK, Shrank WH, Levin RL, et al. Measuring concurrent adherence to multiple related medications. Am J Manag Care 2009; 15: 457–64.

⁵⁶ Matza LS, Park J, Coyne KS, et al. Derivation and validation of the ASK-12 adherence barrier survey. Ann Pharmacother 2009; 43: 1621–30. doi:10.1345/aph.1M174

⁵⁷ Bush K, Kivlahan DR, McDonell MB. The AUDIT alcohol consumption questions (AUDIT-C): an effective brief screening test for problem drinking. ambulatory care quality improvement project (ACQUIP). alcohol use disorders identification test. Arch Intern Med 1998; 158: 1789–95. doi:10.1001/archinte.158.16.1789

⁵⁸ Centers for Disease Control and Prevention. Behavioral risk factor surveillance system questionnaire Atlanta. GA U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2013. Available: http://www.cdc.gov/brfss/questionnaires/pdf-ques/2013%20BRFSS_English.pdf

⁵⁹ Segal-Isaacson CJ, Wylie-Rosett J, Gans KM. Validation of a short dietary assessment questionnaire: the rapid eating and activity assessment for participants short version (REAP-S). Diabetes Educ 2004; 30: 774, 776, 778 passim. doi:10.1177/014572170403000512

⁶⁰ Smith SG, O’Conor R, Aitken W, et al. Disparities in registration and use of an online patient portal among older adults: findings from the Litcog cohort. J Am Med Inform Assoc 2015; 22: 888–95. doi:10.1093/jamia/ocv025

⁶¹ Nelson DE, Kreps GL, Hesse BW, et al. The health information national trends survey (HINTS): development, design, and dissemination. J Health Commun 2004; 9: 443–60; doi:10.1080/10810730490504233

⁶² Fox S, Duggan M. Health online 2013. Washington, DC: Pew Internet & American Life Project, 2013.

⁶³ Polinski JM, Barker T, Gagliano N, et al. Patients' satisfaction with and preference for telehealth visits. J Gen Intern Med 2016; 31: 269–75. doi:10.1007/s11606-015-3489-x

⁶⁴ Baik SH, Fox RS, Mills SD, et al. Reliability and validity of the perceived stress scale-10 in hispanic Americans with English or Spanish language preference. J Health Psychol 2019; 24: 628–39. doi:10.1177/1359105316684938

⁶⁵ Geier TJ, Hunt JC, Hanson JL, et al. Validation of abbreviated four- and eight-item versions of the PTSD checklist for DSM-5 in a traumatically injured sample. J Trauma Stress 2020; 33: 218–26. doi:10.1002/jts.22478

⁶⁶ Hughes ME, Waite LJ, Hawkley LC, et al. A short scale for measuring loneliness in large surveys: results from two population-based studies. Res Aging 2004; 26: 655–72. doi:10.1177/0164027504268574

⁶⁷ Yu L, Buysse DJ, Germain A, et al. Development of short forms from the PROMIS™ sleep disturbance and sleep-related impairment item banks. Behav Sleep Med 2011; 10: 6–24. doi:10.1080/15402002.2012.636266

⁶⁸ Tomaszewski Farias S, Mungas D, Harvey DJ, et al. The measurement of everyday cognition: development and validation of a short form of the everyday cognition scales. Alzheimers Dement 2011; 7: 593–601. doi:10.1016/j.jalz.2011.02.007

⁶⁹ Farias ST, Mungas D, Reed BR, et al. The measurement of everyday cognition (ECog): scale development and psychometric properties. Neuropsychology 2008; 22: 531–44. doi:10.1037/0894-4105.22.4.531

⁷⁰ Cella D, Riley W, Stone A, et al. The patient-reported outcomes measurement information system (PROMIS) developed and tested its first wave of adult self-reported health outcome item banks: 2005–2008. J Clin Epidemiol 2010; 63: 1179–94. doi:10.1016/j.jclinepi.2010.04.011

⁷¹ Topp CW, Østergaard SD, Søndergaard S, et al. The WHO-5 well-being index: a systematic review of the literature. Psychother Psychosom 2015; 84: 167–76. doi:10.1159/000376585

⁷² World Health Organization. Wellbeing Measures in Primary Health Care. The Depcare Project, 1998.

⁷³ George J, Phun Y-T, Bailey MJ, et al. Development and validation of the medication regimen complexity index. Ann Pharmacother 2004; 38: 1369–76. doi:10.1345/aph.1D479

⁷⁴ Jones BL, Nagin DS, Roeder K. A SAS procedure based on mixture models for estimating developmental trajectories. Sociol Methods Res 2001; 29: 374–93. doi:10.1177/0049124101029003005

⁷⁵ Jones BL, Nagin DS. Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol Methods Res 2007; 35: 542–71. doi:10.1177/0049124106292364

⁷⁶ Iacobucci D. Mediation analysis and categorical variables: the final frontier. J Consum Psychol 2012; 22: 582–94. doi:10.1016/j.jcps.2012.03.006

⁷⁷ Preacher KJ, Rucker DD, Hayes AF. Addressing moderated mediation hypotheses: theory, methods, and prescriptions. Multivariate Behav Res 2007; 42: 185–227. doi:10.1080/00273170701341316

⁷⁸ Preacher KJ, Hayes AF. SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behav Res Methods Instrum Comput 2004; 36: 717–31. doi:10.3758/bf03206553