Our study found that, within our sample, the proportion of patients with chronic pain using cannabis was 2.4%. Moreover, over the course of our study, from 2011 to 2015, we identified a significant and progressive increase in the number of patients using cannabis. In patients with chronic pain, cannabis use more than doubled during this period, from 33,189 to 72,114 people. We observed an increasing trend in patients with chronic regional pain syndrome (1.25–1.96%), trauma (5.23–8.98%), spondylosis 0.57–1.2%), FBSS (0.47–0.97%), and other chronic pain diagnoses (2.4–3.49%) (Table 2). This overall increase is not surprising given that several studies have shown cannabis to be effective in mitigating inflammation and demonstrated benefit to chronic pain symptoms with the use of cannabis, including improved pain, functional outcomes, and quality of life in patients with chronic pain syndromes [17–23]. With chronic pain projected to increase over the next 2 decades to a rate of one in three people from the current rate of one in five people, our findings foretell that cannabis use can be projected to increase even more rapidly [24–26]. As evidence of benefits of cannabis use grows, even countries slow to embrace medical cannabis legislation have seen increases in patient awareness of and interest in the therapy [27].

While our study identified a rising prevalence in the use of cannabis among patients with chronic pain, patients with chronic pain frequently have comorbid conditions that may additionally benefit from cannabis use. Various studies have reported that the increased enactment of medicinal cannabis will lead to improved access to cannabis as an alternative pharmaceutical therapy for the treatment of several chronic pain conditions [28, 29]. Across all states that had legalized cannabis for medical use, Baron et al. found pain syndromes, such as migraines, headaches, and arthritis, reported as the most common indication for cannabis use. In addition they found opioids to account for the majority of prescription medications substituted with cannabis, followed by anti-depressants, anti-anxiety medications, and nonsteroidal anti-inflammatory drugs (NSAIDs) [28].

Despite the rising trend of cannabinoid use, which implies improved access to patients who have a need for cannabinoid management of chronic pain, concerns remain regarding cannabis serving as a common illicit gateway drug [29]. This is evidenced by studies that demonstrate a rise in cannabis use among teens, supported by a corresponding decline in the perception of associated risk [30, 31]. In the whole population, approximately 12% of people aged ≥ 12 years had used cannabis at least once in the past year [30]. This observed increase in use suggests a perceived risk of use as well as increase of use among the general population is significant and may contribute to our finding of an increased trend of cannabis use in chronic pain patients. As a whole, these observations may shed light on the overall transformation of knowledge, attitudes, and beliefs in the US population toward the social acceptability of cannabis use; however, further research is needed to investigate outcomes of widespread increases in cannabis use.

Our study showed that the average age of chronic pain patients who use cannabis increased from 42.8 to 45.4 years during the years 2011 to 2015. Moreover, the highest proportion of cannabis use was observed in patients aged 45–84 years and the lowest in those > 85. These results are in line with the findings of Baron et al. who found a mean age of 40 years among all cannabis users but conflict with the results of Vin-Ravin et al., who found 65% of cannabis users to be < 40 years old [28, 32].

The majority of chronic pain patients using cannabis belonged to a lower socioeconomic class (SEC) compared to non-cannabis users. This was similarly shown by Hill et al. and Vin-Raviv et al. [29, 32]. The lowest proportion of cannabis use, 11.9%, was found in the highest income quartile while the highest proportion of cannabis use, 39.4%, was in the lowest income quartile. In our analysis of the payer base, we found that patients with cannabis use were mostly covered by Medicaid. Furthermore, we observed an increased proportion of cannabis users on Medicaid, from 32.7 to 40.4%. Cannabis users were least covered by private insurance, 11.9%, compared to non-cannabis users, 36.8%.

In contrast, Baron et al. reported cannabis use to be associated with a higher SEC [28]. Moreover, Baron et al. found that patients who use cannabis for chronic pain syndromes were more likely to be employed (62.8%) and have education beyond high school (73.6%). Only 5.5% of the patients using medicinal cannabis for chronic pain were retired, which is in line with a decrease in observed use with increasing age. Their study was conducted via electronic questionnaire, which may be a source of selection bias leading to a preponderance of the upper class accounting for the discrepancy between our results. Despite this limitation and potential for bias, a trend of increased cannabis use among patients of higher SEC may be correlated with increased awareness, knowledge, and access to the benefits of cannabis use. Our analysis however does not support this trend; over the course of our study period, the lowest income quartile saw an increase in the proportion of cannabis users, 36.1–40.9%, whereas the highest income quartile saw a decrease in the proportion of cannabis users, 13.1–12.0%. A history of chronic pain may contribute to this observed discrepancy although further studies are needed for elucidation.

Healthcare cost and utilization have been shown to be highest in patients with increasing numbers of comorbidities [33, 34]. In our study we note that the total hospital cost of patients with cannabis use increased from $31,271 (SD $1333) in 2011 to $38,684 (SD $946) in 2015 (Table 3), with the mean length of hospital stay also increasing from 2012 to 2015. Interestingly, in an analysis of cannabis use and inpatient outcomes among hospitalized patients, Vin-Ravin et al. reported cannabis users as having fewer comorbid conditions on admission (78% vs. 71%) compared to non-users [32]. This led Vin-Ravin et al. to report a significantly reduced odds of in-hospital mortality and morbidity in cannabis users compared with non-cannabis users (odds ratios of 0.41 and 0.44, respectively). This indicates that cannabis use may have some degree of all cause morbidity and mortality benefit among hospitalized patients. The increase in healthcare spending among chronic pain patients using cannabis is difficult to explain based on these previous studies together. We would expect increases in cannabis use to lead to decreases in overall spending; however, our findings show the opposite trend. The disparity in our findings may partly be due to the fact that the majority of cannabis users in the study by Vin-Ravin et al. were < 40 years old, who are less likely to have comorbidities and ultimately have lower health care costs compared with the predominant cannabis users in our study who were between 45 and 64 years old. Moreover, the increased spending among chronic pain patients using cannabis, which we note in our study, may largely be due to an overall trend of increased hospital spending. Our findings are supported by the concept of “super-additive costs” per Cortaredona et al. and may be explained by an increasing trend of older patients having more comorbidities and as a result higher health care utilization and resultant costs [34].

Our study has several limitations. First, the NIS database does not contain variables that can evaluate disease severity. The degree of cannabis use and its relationship to worsening chronic pain was not available for analysis. In addition, the definite relationship between chronic pain and cannabis may not be well established with this administrative data set. We included only chronic pain patients with a history of cannabis use. However, it is very difficult to elucidate the indication for cannabis prescription or consumption among our chronic pain cohort. Cannabis use may be secondary to other medical comorbidity outside of chronic pain. Second, the NIS database does not include detailed comorbid cannabis use disorder-related cost information, but rather total hospital charges. This unfortunately makes specific cost analysis, such as intensive care unit (ICU) admission cost, cost of administered anesthesia, procedure room charges, and postoperative charges, impossible. Third, healthcare utilization outcomes available for analysis in this study may not be directly related to comorbid cannabis use. They could be influenced by multiple factors such as payer status and social factors. Fourth, the NIS database is highly dependent on the selection and accuracy of the appropriate diagnostic ICD codes reported [35]. The rise in concomitant cannabis use may reflect a coding artifact or the implementation of new ICD-9 codes. This makes it is very difficult to determine the actual incidence of cannabis use in hospitalized patients with chronic pain. Furthermore, the criteria used for diagnosis of conditions such as CRPS were not reported in our database. The lack of standardization may affect the quality of our result.