This is a secondary analysis of cross-sectional survey data conducted among adults (≥18 years old) in US states with legal medical/recreational cannabis or Canada who were self-administering cannabis for chronic pain symptoms. Cannabis dispensaries and certification clinics shared an anonymous Qualtrics (Provo, UT) survey link with their client databases and on social media between January and August 2018. Participants freely consented to participate, could drop out at any time, and were not compensated. We descriptively analyzed open-ended written responses to the following question, “Please describe your typical routine for using cannabis. For example, ‘I vaporize a high CBD cannabis distillate for pain relief throughout the day, and then take 10 drops of a strong THC tincture before bed.’” As with previous studies with this cohort (Boehnke et al. 2019b; Boehnke et al. 2019c), analyses initially included 1321 participants from a pool of 1697 responses (77.8% completion rate of the original survey). In the current study, participants were excluded if they either did not respond to the question (n=68, 5.1%), their responses were vague (e.g., “I take a maintenance dose”, n=48, 3.6%,) or they did not mention an administration route (n=118, 8.9%), resulting in 1087 responses (64.1% of the original N=1697 responses). Study procedures were approved as an exempt study (HUM00079274) by the Institutional Review Board at the University of Michigan due to the anonymous, confidential survey study design of this project.
Qualitative descriptive analysis of routine data: Codebook creation
We created the codebook using an inductive descriptive approach whereby codes were developed based on respondent answers to identify variation in cannabis use routines (Sandelowski 2000). Descriptive analysis is a strategy to summarize participant experiences, focusing on their own words and how they relay their experiences. Codebook development involved two members of the research team (K.B., J.M.) reading through a sample of responses, independently developing a list of codes that described participant responses, and collaboratively developing a first draft of the codebook. The codebook included the list of codes, definitions of responses that fit each code, and methods for resolving discrepancies. Next, L.Y. coded the entire dataset, updating the codebook as necessary and meeting with the research team to discuss adjustments. After coding, we validated for inter-coder reliability. Ten percent of the entire data-set was selected for validation (Mao 2017) using the random number generator feature in Microsoft Excel, which K.B. then independently checked to confirm application of codes. After validation, we met to discuss discrepancies and resolve any responses that had created ambiguity. Our final codebook included multiple elements of use routines, including cannabinoid content, cannabis subtype, administration route, and timing. Brief descriptions of each component are below, and full descriptions may be found in the codebook (see Additional file 1: Appendix 1).
Administration routes included smoking, vaporizing, edibles, tinctures, topicals, and other. “Other” incorporated unspecified or rare administration routes and were not included in these categories.
Timing of use was coded as: Morning (before 12 PM); Afternoon (12–5 PM); Night (after 5 PM); and throughout the day (e.g., “every few hours,” “all the time”). We also created two non-specific codes: as needed and uncertain. As needed was used when language indicated that use is not routine but related to a specific symptom (e.g., “when I have spasms”). Uncertain was used when timing remained unclear (e.g., “I smoke”).
Cannabinoid content refers to CBD, THC, or the combination. When participants did not mention the type of cannabinoid, these responses were coded as “unknown cannabinoids.”
Cannabis subtype refers to indica, sativa, or hybrid cannabis.
We created subgroups based on mutually exclusive categories of inhalation (smoking, vaping) and non-inhalation (edibles, tinctures, topicals). These groupings were chosen because effect onset is known to be quite different between administration routes (e.g., 5–10 min for smoking vs. 1–3 h for edibles), which may result in different effects and signify different use populations (MacCallum and Russo 2018). Indeed, inhalation routes are most commonly used and associated with more addiction/abuse potential and health risks but also more effective pain relief (MacCallum and Russo 2018; Andreae et al. 2015), while non-inhalation routes are used more by newer medical cannabis patients (Boehnke et al. 2019b).
Measures for exploratory analysis
Demographic and clinical characteristics
We collected data on sex, age, socioeconomic status, concomitant self-reported current opioid and benzodiazepine use, alcohol use (never vs. ever drank), frequency of cannabis use (days/week, times/day), and cigarette use (never, former, and current use).
Changes in pain and health
We asked participants how their health and pain had changed since they started using medical cannabis, with response options on a five-point Likert scale from “Declined a lot” to “Improved a lot” for health and “Increased a lot” to “Decreased a lot” for pain. Responses for both pain and health were analyzed continuously, with − 2 indicating health declining a lot or pain increasing a lot and 2 indicating that health improved a lot or pain decreased a lot.
As described previously (Boehnke et al. 2019c), participants reported classes of pain medication for which they substituted medical cannabis, including opioids, benzodiazepines, gabapentanoids, disease-modifying anti-rheumatic drugs, non-steroidal anti-inflammatory drugs, selective serotonin reuptake inhibitors, selective norepinephrine reuptake inhibitors, and other. The number of substituted medication classes was counted and analyzed continuously.
Qualitative analyses are presented descriptively using representative quotes and counts of different cannabis use routine variables: e.g., administration routes, cannabinoid content, cannabis subtype, use timing. After subgrouping routines by administration routes (non-inhalation, inhalation, non-inhalation + inhalation), we explored differences for categorical variables (e.g., sex, income, tobacco, alcohol) using Pearson’s chi-square. For continuous variables (age, perceived changes in pain/health, substitutions), we assessed differences using one-way analysis of variance (ANOVA) followed by post hoc testing with Tukey’s test to identify subgroups driving significant differences of changes in pain, health, and substitution. Continuous variables were assessed for normality and were normally distributed, and are reported as mean +/− standard deviation. All tests were two-tailed, and significance set at p < 0.05. Analyses were performed using SPSS 25 (IBM, Armonk, New York) and Excel version 2019 (Microsoft, Redmond, Washington).