A Systematic Review of Pain Management Education in Graduate Medical Education

One of the few pathologies shared among most medical specialties is pain: 20% to 50% of primary care patients present to clinic with chronic non-cancer pain,^1,2  40% to 60% of emergency department visits are due to a pain-related chief complaint,^3,4  and at least 80% of surgical patients have some degree of postoperative pain.^5,6  Poor control of these painful syndromes is associated with poor patient outcomes, high morbidity, and increased costs to society.^7-9  However, despite the clear importance of effectively managing pain, physician knowledge and comfort around pain management is alarmingly low.^10,11 

Acute and chronic non-cancer pain management requires unique approaches that balance benefits of analgesia with potential risks, including opioid use disorder.¹²  One of the most highlighted issues of the last decade is the opioid epidemic, which is primarily a concern regarding opioid use disorder in patients without cancer pain.^13,14  One cited contributing factor to the opioid epidemic is physician overprescription,^15,16  which may, in part, be addressed through physician education.^17,18  In recognition of the connection between undereducation, poor management, and overprescription, states have begun to mandate additional training in pain management as part of licensing requirements.¹⁹  In spite of these measures, a recent review of pain education in medical schools noted that US schools dedicate a median of only 9 hours to pain management, which has been critiqued as insufficient for addressing societal needs.^20,21  Within graduate medical education (GME) programs, residents have reported feeling uncomfortable and unprepared to manage pain, at least partially due to undereducation.^22-24 

In this study, we sought to examine pain management education within GME programs. Specifically, we aimed to (1) evaluate and appraise the literature on acute and chronic non-cancer pain management education as it currently exists in GME programs through an analysis of methodology and educational outcomes, and (2) summarize the educational methods and foci described in said literature.

Because the goal of our review was to assess and summarize research on pain management education in GME, we chose a systematic review, as this design allows for systematic identification and evaluation of all available literature. This study was executed in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).²⁵ 

We searched PubMed, MedEdPORTAL, and ERIC (Education Resources Information Center) using variable terms to identify English-language articles published online before March 2019 that focused on educational interventions around acute and chronic pain management in GME. Our search terms included iterations and synonyms of our population (eg, residents, fellows, interns); intervention (eg, education, curricula, didactic); and topic (eg, pain management, opioids, analgesia). While the Population, Intervention, Comparison, Outcome (PICO) framework is commonly employed to structure a search strategy, we did not target a specific comparison group or outcome in order to keep our search as broad as possible. See the online supplementary data for full study protocol, including complete search terms.

In order to meet inclusion criteria for our review, studies had to have:

1. A majority focus on non-cancer pain management education, defined as an intervention that focused ≥50% of time or content (as gathered from the studies) on pain management physiology, pharmacology, prescribing habits, or patient communication strategies.

2. A learner base that was composed of a majority of GME learners, defined as a learner base with ≥50% medical residents and/or fellows.

3. A structured, formal, educational intervention defined as a didactic session, group discussion, simulation, online module, written instruction, or tool.

4. A reported outcome, defined as any structured evaluation of the implemented intervention.

Given the heterogeneity of educational studies targeting pain management, inclusion criteria were designed to allow for the broadest search possible. We chose to include only articles with outcomes in order to target interventions that had been implemented and studied as opposed to interventions that had only been crafted or proposed. Articles were excluded if they had a primary focus on cancer-related, end-of-life, or hospice pain management, as these categories of pain management weigh the risks of opioid use and opioid use disorder differently than in acute and chronic non-cancer pain management. We also excluded articles primarily focused on interventional and procedural analgesic techniques or substance use disorders as these fell outside the aims of the paper. Other reasons for exclusions were non-English-language articles, opinion pieces, and non-peer-reviewed literature.

Two study authors (Z.M., K.T.) independently reviewed all titles and abstracts and saved relevant studies using Zotero version 5.0.87 (Corporation for Digital Scholarship, Vienna, Virginia). In cases of disagreement, the full text was independently reviewed by both authors and reconsidered for inclusion. If the 2 authors continued to disagree, the full text was rereviewed, the article discussed, and consensus agreement was reached based on the predetermined inclusion/exclusion criteria. In cases of continued disagreement, a third and more senior author (A.P.) served as an arbitrator.

Study data were managed using REDCap version 9.5.24 hosted at the University of Chicago (Vanderbilt University, Nashville, Tennessee). We created a data extraction form (online supplementary data) that allowed for analysis based on the BEME (Best Evidence Medical Education) Collaboration²⁶ ; this was piloted with 10 articles before use. Data extracted from the articles consisted of 3 domains: (1) Journal (name, Journal Citation Reports Impact Factor, SCImago Journal Rank); (2) Study characteristics (publication year, number, name, and location of participating institution[s]); and (3) Participant characteristics (number, level of training, and medical specialty).

One author (Z.M.) with expertise in medical education also collected educational characteristics of the included studies by examining methods, images, tables, and appendices. These data were then categorized by method and educational focus (online supplementary data). Given the nuances of this content, only the reviewer with a strong understanding of educational methods conducted this aspect of the review.

We assessed methodological quality using 2 separate instruments: the Medical Education Research Study Quality Instrument (MERSQI) and the Newcastle-Ottawa Scale–Education (NOS-E).

The MERSQI was designed to “measure the [methodological] quality of experimental, quasi-experimental, and observational studies.”²⁷  The MERSQI is composed of 10 items reflecting 6 domains of research quality (study design, sampling, type of data, validity, data analysis, outcomes); it has demonstrated criterion validity as well as high inter- and intra-rater reliability across items.²⁸ 

The NOS-E was developed for “use in a meta-analysis of Internet-based education for health professionals” by modifying the Newcastle-Ottawa Scale, an instrument developed to assess the quality of comparative nonrandomized studies included in meta-analyses.^29,30  The NOS-E evaluates 5 items: representativeness of the intervention group, selection of the comparison group, comparability of the groups, study retention, and blinding. NOS-E authors demonstrated high inter-rater reliability and as well as correlation with MERSQI scores.²⁹ 

The MERSQI and NOS-E are both designed to score methodological quality. The MERSQI, however, is slightly more objective and focuses on design, whereas the NOS-E focuses on procedure. The present study follows the suggestion by Cook et al that these tools should serve to complement one another in analysis of methodology.²⁹ 

Two authors (Z.M., K.T.) independently scored each article using both the MERSQI and the NOS-E; these scores were used to calculate an initial intraclass correlation coefficient (ICC) between the 2 raters to evaluate reliability. Disagreements were resolved by consensus. Descriptive statistics, total scores, and subscale scores of the MERSQI and NOS-E were calculated using Excel version 16.16.21 (Microsoft Corp, Redmond, Washington) and R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria).

The initial query yielded 6149 articles. Following our prespecified review process, we refined the search to 134 articles. These 134 articles were reviewed in full and resulted in 26 articles that met inclusion criteria (Figure).^31-56  A senior author (A.P.) served as an arbitrator to determine inclusion eligibility for 6 articles in the final stage of review.

Included articles were published between 1996 and 2019; 24 studies (92%) were published after 2000 and 17 (65%) published after 2009 (Table 1). One included study was published as a presentation given at a national conference.⁴⁵  The rest were published in journals that ranged from regional to international and were related to GME, pain, and various specialties. The mean journal impact factor was 2.61 (SD 1.31) and the mean SCImago Journal Rank was 0.96 (SD 0.52). The most commonly represented fields were internal medicine (n=17, 65%), family medicine (n=4, 15%), pediatrics (n=4, 15%), surgical subspecialties (n=4, 15%), emergency medicine (n=3, 12%), and anesthesia (n=3, 12%). The majority of scholarship came from within the United States (n=24, 92%) and from institutions with a university affiliation (n=23, 88%). While our search terms were intended to capture all interventions within GME, we found only one study that included fellows.⁴⁸  Only 2 papers (8%) described interventions delivered to faculty as well as residents.^39,43 

Our review examined articles that described educational interventions directed toward acute and/or chronic non-cancer pain management. The majority (n=15, 58%) of included studies were focused solely on chronic pain management. Seven (27%) focused on acute pain and 3 (12%) had educational elements of both. There was one study (4%) that did not specify whether their education was directed toward acute or chronic pain management. Intervention lengths were highly variable across studies: the shortest intervention was a single 30-minute lecture, while the longest intervention was spread out over 1.5 years. The minority (n=7, 27%) took place over 1 month or longer. The most common educational methods were traditional lectures (n=16, 62%) and case-based learning (n=14, 54%). Four studies (15%) included use of pocket cards, and 3 studies (12%) provided a general resource kit. Four studies (17%) used standardized patients or OSCEs in their intervention; 3 (13%) incorporated small group learning. The majority (n=16, 62%) of studies implemented multiple methods.

The mean consensus MERSQI score was 11.6 (SD 2.33) out of a maximum of 18 (Table 2). Reliability for MERSQI scores was high (ICC=0.94, 95% CI 0.87-0.97). The most prevalent study design was a single group, pretest-posttest design (n=16, 64%). The majority of studies sampled from single institutions (n=21, 84%). Forty-eight percent (n=12) of articles had response rates greater than 75%. Some articles included content validity evidence for their evaluation instrument (n=12, 48%); however, few included internal structure (n=5, 20%) or relationships to other variables (n=3, 12%). Data analysis was appropriate in nearly all studies (n=24, 97%) and went beyond simple descriptive analyses in most (n=21, 84%). The most common study outcomes were post-intervention test scores, which evaluated effects of the intervention on learner knowledge (n=12, 48%), and post-intervention surveys, which examined learner satisfaction, perceptions, and/or attitudes (n=6, 24%). Five studies (20%) evaluated learner behaviors; most of these looked at changes in resident prescribing behavior. Only 2 studies (8%) evaluated patient or health care–centered outcomes; both tracked changes in patient pain scores before and after their interventions.

The mean consensus NOS-E score was 2.60 (SD 1.22) out of a maximum of 6 (Table 3); reviewers had an ICC of 0.93 (95% CI 0.83-0.97). The NOS-E representativeness domain is similar to the response rate domain of the MERSQI; as expected, the findings here are similar to the MERSQI with the majority (n=15, 60%) of studies having an intervention group that was very or somewhat representative of the average learner in the community. Few studies (n=4, 16%) had a comparison group from the same community of participants; the majority did not have a separate comparison group (n=19, 76%). Studies largely had retention rates that were unlikely to introduce bias per NOS-E definitions (n=21, 84%). The outcomes domain of the NOS-E is solely concerned with blinding; the majority of outcomes assessments were blinded (n=18, 72%). We were unable to calculate MERSQI and NOS-E scores for one included paper because it was published without sufficient detail (Table 4).³⁴ 

In this systematic review of acute and chronic non-cancer pain management education within GME, we found room for improvement largely related to study design and methodology. Most studies were conducted at a single site and assessed their interventions using outcomes at low Kirkpatrick levels.⁵⁷  We also noted a preponderance of brief educational interventions mostly focused on chronic pain management.

As Cook and Reed suggest, insight into study quality can be attained by examining MERSQI and NOS-E domain scores.²⁹  Specific domains of weakness noted were design, sampling, use of validity evidence, and outcomes. The majority of papers included in this review used a single group, pretest-posttest design. Only Sullivan et al used a multicenter randomized control design and resultingly had the highest total MERQSI score.⁴²  Multi-institutional studies, though logistically harder, would strengthen study results. One way to ease the burden of performing multi-institutional studies is through the creation of collaborative research groups. One successful example of this is the Emergency Medicine Education Research Alliance, a network of researchers that collaborate to produce high-quality, multi-institutional research.⁵⁸ 

The outcomes domain of the MERSQI ties directly to Kirkpatrick's hierarchy for evaluation; the majority of the included studies assessed either learner reactions (level 1) or knowledge (level 2), but rarely learner behavior (level 3) or patient/health care outcomes (level 4).⁵⁷  While this is an important finding and critique, it is not uncommon in the field of medical education; other reviews of education from various fields have made similar observations.^59-62  Some have posited that these restricted outcomes measures are due to a lack of external funding for medical education research²⁷ ; however, others argue that medical education researchers have the responsibility to aggressively pursue funding in order to evaluate high-level outcomes and push the field forward.⁶³  It should be noted that, while higher Kirkpatrick's levels may seem inherently better as outcomes, low-level outcomes are still appropriate and useful in many contexts (ie, program evaluation, novel interventions).⁶⁴ 

Despite our inclusive search for both acute and chronic non-cancer pain management education, we found few interventions that focused on acute pain. While this undoubtedly reflects the urgency to address opioid utilization for chronic pain conditions, we feel that acute pain management education is equally important for patients and clinicians, especially in light of recent literature suggesting the mismanagement of acute pain may lead to long-term dependence and poor outcomes.^15,65-69  While it was outside the scope of our review, we did note that a number of articles were excluded due to a focus on cancer-related pain or end-of-life care. Some principles of acute pain management may have been included in these articles.

The educational methods described in articles included in this review are encouraging. In choosing educational methods, Thomas et al offers 3 guidelines: (1) maintain congruence between objective and methods; (2) use multiple methods; and (3) choose methods that are feasible in terms of resources.⁷⁰  Most included papers focused on cognitive objectives (ie, build the knowledge base of learners). Lectures and case-based learning, which were the most commonly employed educational methods, are well-suited to address these objectives.⁷⁰  However, there are certainly behavioral and communicative aspects to pain management which are best targeted by supervised clinical experiences, role-playing, and the use of standardized patients⁷⁰ ; these methods were employed by just 7 (29%) studies.

The timeline of the studies in this review are also of interest. While analgesia has always been an important aspect of patient care, pedagogical interest in this topic seems to have increased recently as a majority of papers included in our analysis were published after 2009. While this may be due to a number of factors, a key contributing factor may be the opioid epidemic: in the late 2000s, literature questioning the role of opioids in pain control began emerging and Purdue Pharma plead guilty to misbranding OxyContin.^71,72  Another possible explanation for the recent surge in publications is the increasing emphasis on evidence-based medical education. Lim and Golub noted that the number of systematic reviews within GME increased from 2 between 1966 and 1974 to 373 between 1995 and 2004.⁷³  While medical education research has existed for some time, over the last decades this field has been experiencing what one author called a “renaissance;” the growing literature on pain management education might be due in part to this evolution.⁷⁴ 

The geographic distribution of included studies in our review is also interesting. Medical education research is largely concentrated in Europe, North America, and Australia; however, we found only 2 non-US studies.⁷⁵  This may be related to the mentioned opioid epidemic concentrated in the United States, but it may also reflect a limitation of our search strategy in that we neglected to include a large open access database run by the Association for Medical Education in Europe called MedEdPublish.

This study has several additional limitations. First, our literature search was limited to 3 databases. Though this permitted an extensive review, future studies could include additional databases and specifically should be inclusive of non-US-based databases. Further, our use of quantitative tools (ie, MERSQI, NOS-E) to assess methodological quality could have overlooked notable qualitative characteristics of study methods. However, we did not explicitly exclude qualitative work, and none of the articles included in our review were purely qualitative. There were a few papers that described curricular innovations and needs assessments in regard to pain management education, neither of which were included here but may offer insight into the state of pain management education.

Given the severity of the problem of undereducation around pain management, the ongoing opioid epidemic, and increased focus in methodologically sound medical education research, we call for more robust studies that might guide field-specific education. While the present study summarized educational methods, future work could evaluate which methods most effectively improve acute and chronic non-cancer pain management in GME. Additionally, future research on pain management in GME should strive to use more rigorous design, multi-institutional and/or interdepartmental sampling, and to target outcomes related to learner behavior or patient outcomes.

In this systematic review of acute and chronic non-cancer pain management education in GME, we found that a majority of studies evaluated chronic pain management educational interventions at single sites through an assessment of learner reactions or knowledge. Notably, despite pain management being a pervasive issue that is not restricted to internal medicine, there were few interventions from other fields.

The authors would like to thank Debra Werner for her support in executing our search.