Educational Intervention for Effective Postoperative Pain Management in Low Resource Settings: Evidence from Ethiopia

The annual number of surgical operations performed is increasing throughout the world. With this rise in the number of surgeries performed, so too, the challenge of effectively managing postoperative pain. Healthcare professionals and patients in education might help in controlling postoperative pain effectively. However, data from low-income countries investigating the impact of educational intervention on postoperative pain are very scanty, and reports from the developed settings are inconclusive. The study has investigated the impact of preoperative patient education and health care professionals education on improving the quality of postoperative pain management; in patients scheduled for major elective orthopedic, gynecologic and general surgery; as measured by patient-reported outcomes.

to reduce postoperative pain [11]. Following this many more studies in the field tried to replicate the findings. However, some authors reported no effect of preoperative patient education after conducting a well-randomized controlled trial [12]; while others claim a positive effect on postoperative pain. Lately, however, the argument whether preoperative patient education is effective or not is starting to materialize in literature [13]. A systematic review and meta-analysis of RCT on the topic also failed to bring consistent results; while some support [14] and others not [15]. In some studies, patients received various forms of other therapeutic interventions along with the preoperative education; authors call this adjunct treatment [16]. Popular adjunct treatments that accompanied preoperative education included either exercise or relaxation [16]. As distinguished from studies performed until now, we have accompanied the preoperative patient education by Healthcare Providers (HCPs') education; by hypothesizing a well educated patient and HCP can perform better synergistically. A review of literature recommends educational programs to include interdisciplinary professionals and policymakers in addition to patients [17].
The aim of this study was to investigate the impact of preoperative patient education and HCPs education on postoperative pain outcome in patients scheduled for major elective orthopedic, gynecologic and general surgery. In addition to testing the impact of the intervention on the outcome; unlike to the previous studies, we have attempted to explain the underlying mechanisms of the preoperative education. Hitherto, it is believed that providing patients with education is to address preoperative level of anxiety in the preoperative period and then subsequently to decrease pain intensity [16]. All experimental studies so far tested whether the education is effective or not in decreasing patient pain intensity and no study attempted to verify if, in fact, preoperative education affects patient postoperative pain through anxiety or any other pathway. One of the emerging statistical tools to understand the causal mechanism behind an intervention is causal mediation analysis [18]. With the help of this method, we have attempted to answer whether the designed intervention was effective and also explain the underlying mechanism. So, this study provides information regarding the effectiveness of preoperative patient education on postoperative pain and also the causal mechanism.

Materials And Methods
This was a quasi-experimental, non-equivalent control group design to assess the effectiveness of educational intervention given to patients and HCPs. The study was   Figure 1 with participants flowchart.

Intervention
The contents of all training materials were based on literature review [13,16,17,19] IASP recommendations [20], international recommendations for low resource settings [21,22] national guidelines [23] and considering the principles of learning sciences. The latter takes into account the conditions, processes, and outcomes of learning [24]. Before the main preoperative patient education, staff members of the intervention hospitals were trained on the effective management of post-surgical pain. Health care providers including those who assume a leading or managerial position, surgeons, gynecologists, nurses, and physiotherapists were invited to participate. A total of 13 participants (3 anesthetists, 3 surgeons, 2 gynecologists, 1 physiotherapist and 4 nurses) attended the workshop. For about 13 hours over 3 days, they were trained on topics related to the obstacles to pain management in low-resource settings, the importance of pain assessment, measurement, and tools, use and application of non-pharmacological methods of pain management. In addition to the theoretical lectures, participants were exposed to the practice sessions.
The hands-on sessions focused on the use of non-pharmacological methods of pain management with emphasis on acupuncture.
After the HCPs education was completed, the night before the surgery, a project team member an anesthetist conducted a one-on-one education verbally. Each session lasted for 15 minutes. Voluntary relatives also attended. In brief, each educational session consists of information regarding why managing postoperative pain is important, nonpharmacological options of pain management, how to take pain medication as directed, report side effects early, participate in the choice of the management of pain with HCPs.
Patients were informed that they should not be shy and always be active in the management of their pain. They were also told how to describe their pain using the pain intensity scales. During the educational session, patients were encouraged to ask questions. Once the question and answer session was completed, each patient was asked to repeat what they have learned. Finally, five questions were asked of all patients and if a patient misses one question education was repeated again. Patients in the control hospitals received care as per usual with no preoperative education. Preoperative education or information for postoperative pain is not part of the care in the setting yet.  [19] Patient role in the management T h e patient should ask for analgesics and insist if the provider is not responding, Patients should not be passive participate in decision [8, 25, 26 , 27]. How to take med directed, manage side-effect early, avoid misconception side-effect early [29] communicate your pain using instruments [30] How to be relaxed and avoid fear prior to surgery Should believe that unmanaged pain is very harmful [19] Available options for treatment Both pharmacological and non-pharmacological methods How to reduce anxiety using various alternatives [31] The night before the operation, the trained data collectors from all sites approached potentially eligible patients. Next, patients were provided with a detailed information sheet describing the study and their potential involvement. For the majority of the patients (who were illiterate), the written information sheet was read to them aloud. But, in any case, a witness's signature was obtained. Patients who provided written (fingerprint in case of the illiterate) and verbal consent to participate completed an intervieweradministered baseline questionnaire, prior to the intervention Patients were aware of their participation in a study but were blind to which condition the hospital was allocated. However, those who administer the interventions and those assessing outcomes were not blind to study allocation. All evaluations were conducted in the participants' surgical wards using a paper and interviewer-administered questionnaire.
Data collectors were different from those who administered the intervention. During the intervention period, however, in addition to the offer to participate in the intervieweradministered self-reported measures, patients in the treatment group were also given additional information about the planned preoperative education. When the patient agreed to participate, a consent was obtained the same way as described above and the preoperative individualized patient education was conducted. Consequently, using the interviewer-administered questionnaire, patient-reported outcomes were collected after the operation at the four-time points explained above.

Outcome measures
The main outcome measure was measured using the Numeric Rating Scale (NRS 0-10; numeric rating scale, but also binary items are included. Patient worst, least and current pain intensity was measured as (NRS 0 = "no pain"-10 = "worst pain possible." The percentage of time the patient spent in severe pain since surgery was measured as 0% = "never in severe pain"-100% = "always in severe pain." Pain interference was measured as functional disability due to pain (NRS 0 = "did not interfere"-10 = "completely interfered"), anxiety and helplessness caused by pain (NRS 0 = "not at all"-10 = "extremely"). Patient perception of care was measured as the degree of pain relief through pain treatment (NRS: 0% = "no relief"-100% = "complete relief"). Patients wish for more analgesics were recorded as binary ("yes or no") answers. Satisfaction with the results of pain treatment was measured with NRS 0 = "extremely dissatisfied"-10 = "extremely satisfied." The original English version has been translated (forward and backward) into two local languages and pilot tested in five steps as per international guidelines [33]. The final version was approved by expert panel to make sure content and face validity. In addition, we have measured the adequacy of pain management using the Pain Management Index (PMI). The index is calculated by first categorizing patients worst pain intensity into 0 (no pain), 1 (1-3: mild pain), 2 (4-6: moderate pain), and 3 (7-10: severe pain). The final score is then subtracted from the strength of analgesic prescribed: which is 0 (no analgesic drug), 1 (non-opioids), 2 (weak opioids), and 3 (strong opioids).
The final score is between -3 to +3, and negative scores inform inadequate treatment.
Originally this was designed to assess the adequacy of cancer pain management; however, its application in surgical patients have been reported [34].

Covariates
The covariates considered in this study were the following: time (since surgery), patient's age and sex, pre-existing chronic pain, patient's physical condition and operating time.
We have also retrieved demographics, medical history information, type of surgery, duration of surgery, type of anesthesia and pain treatment from the medical records.

Statistical analysis
Before testing the effectiveness of the intervention, we calculated mean and SD for normally distributed continuous variables and medians (min, max) in case of skewed distributions. Categorical variables were summarized as numbers (percent). In order to assess the influence of selection bias, differences in baseline clinical and demographic variables at the baseline were evaluated using univariate generalized linear models and using Chi-square test. Comparison of changes in the outcomes of interest over time between the control and treatment group were analyzed using linear mixed effect model.
Covariates in the final model were selected using backward elimination, which begins with the maximum full model and then deleting variables of no value. However, age, sex, type of surgery, and chronic pain severity were left in the model despite statistical insignificance, to avoid omitting a significant variable (avoid any Type II errors) and therefore maximize validity and predictive power, which is a good practice [35]. To adjust for baseline differences between the treatment and control groups, in addition to the baseline outcome value, we have also added additional fixed terms to the LME model.
These include age, sex, type and duration of surgery, type of anesthesia, ASAclassification and preoperative chronic pain severity. In the final model, two random effects were included: the random intercept and the random slope of time (that is, number of hours after the operation) for patients.
Sensitivity analysis.
To test how robust our findings are we conducted a sensitivity analyses using propensity score analysis [36,37]. Robins et al., in 1994, proposed the doubly robust (DR) estimator, which is an amendment of the IPW methods [38]. This method brings together both the outcome regression model and the propensity score model. For this reason, the investigator has two opportunities (chances) of specifying the model correctly. Even if either the propensity score model of the outcome regression model misspecified, the DR remains consistent [38 , 39]. A proper weighting and correct propensity score model removes confounding with respect to the measured baseline covariates, and thus, the average treatment effects obtained reflect the true population average [40]. Using a DR approach can compensate for a lack of covariate balance, unlike to other matching techniques of the propensity score. Moreover, with other previously mentioned matching techniques, the dataset can be pre-processed by "trimming" away (removing) individuals with extreme PS, while attempting balance [41]. Therefore, this method of estimation was used in this particular study, to calculate the average treatment effect.

Causal Mediation analysis
In the present study we have implemented the within-subject 1-1-1 multilevel mediation, also known as lower level mediation [42], page 179]. In longitudinal, within-subject mediation, X, M, and Y can vary either within-subjects (level-1), between-subjects (level-2), or both [42]. Krull and MacKinnon outlined three specific multilevel mediation scenarios: 2 → 2 → 1, 2 → 1 → 1, and 1 → 1 → 1 [43]. Since the mediator (patient participation in decision making) is a level-1 variable and the treatment exposure was also individualized patient education, which is also a level-1 exposure and the outcome variable is also measured at level 1 (patients' worst pain intensity), we have conducted a 1 → 1 → 1, within-subject mediation. We have followed the procedure described by Bolger and Laurenceau [42]. We have performed 1000 sample bootstrap procedure to estimate 95% confidence intervals (CIs) to test the significance of indirect links and CIs are expected not to contain 0 and only then the indirect links are considered to be significant [44]. The mediation analysis was also adjusted for all measured baseline confounders.
The treatment condition (treated vs control) is represented by X, patients participation in decision making is labeled M, and patients' rating of worst pain intensity (patients' satisfaction for the second mediation model) is labeled Y. The total effect was calculated using the formula from Kenny, Korchmaros, and Bolger [45] which is given by: Here we see that c, the relationship between X and Y for the typical patient, is equal to the sum of (1) ab, the product of the X-to-M and the M-to-Y coefficients for the typical subject; (2) c', the coefficient representing the unmediated portion of the X-to-Y relationship for the typical subject; and (3) sajbj, the covariance of between-subjects differences in the X-to-M and M-to-Y relationships. Including the final covariance term () is very important in multilevel mediation and it has an important implications for estimates of mediated effects. It represents that the extent that those patients whose participation in decision making score is most affected by the treatment are the same patients whose pain intensity (patients' satisfaction for model 2) is most affected by their participation in decision making, then the overall mediated effect will be greater than one would expect from the ab product alone [42]. All data management, linear mixed model building, and propensity score weighting was done using STATA "perceptions of care" (r=.62). All the above parameters were consistent and very much comparable with the reports of the original authors [37], except for the four-factor solution where the original authors reported 3-factor structure. However, the phase-one data of the original authors reported four-factor solution with a total explained variance of 60.78% [37, page=1368], which is consistent with our findings. As it is a common practice in the field to do so [37], discriminant validity was assessed by comparing surgical category of patients. We used Mann-Whitney U tests and chi-square tests to contrast groups. Because of a small proportion of orthopedic and gynecologic patients, the two were combined together and contrasted with the general surgical patients . Except for   least pain intensity, pain interference with sleeping, pain interfering with activities out (Table 1).   after the intervention. However, the proportion of patients inadequately treated increased from 30% to 41% and from 1% to 23% at the 24 and 48 hours after the surgery respectively.
The same trend was observed in the control group that patients inadequately treated   Observing significant effects at later postoperative periods compared to the early timepoints could arise from the natural surgical ward contexts in the low resource settings, the nature of preoperative information itself and complex psychological phenomena.
There is a limit to what extent pain management can be successful without the use of strong analgesics. No matter how effective an education is, it is an adjunct treatment [19] and can not replace effective analgesics. At the time of this study, no opioids were available for the surgical patient and Ethiopia is classified as a country with nil morphine per capita [48]. Also, giving patients specific information about the importance of good postoperative analgesia might improve their understanding, however, this does not translate necessarily to better postoperative pain outcome. Psychologists explain this by the difference between automatic and planned behavior [49]. Automatic processes, or habits, enable behaviors to be carried out with a little or no demand for cognitive effort, and they make behavioral changes very complicated [50]. Education, therefore, can lead to improved knowledge; however, this does not necessarily change old beliefs and habits.
And it might be possible that patients can have increased knowledge of pain treatment and increases participation, without the desired changes in their beliefs or behaviors in accepting analgesics after surgery [50]. The results of this study should encourage HCPs, or researcher that even without opioids with education and non-pharmacological options of pain management, this study demonstrated that improvement can be achieved at least after 12 hours of the surgery.
The difference between patients' worst level of pain with that of current level of pain and, least level of pain, could be associated with the fact that these intensity measures (least and current) are not as sensitive as worst pain intensity in detecting treatment effects, and authors have been recommending against [51]. A clinical trial in Taiwan also reported no effect of the treatment when the outcome was current level of pain and the average level of pain, instead of worst pain intensity [52]. It is also worthy to mention that a recent RCT from Germany, reported no superiority of preoperative patient education over the standard of care for most of the outcome measures authors used, including postoperative pain intensity [53]. Patients' participation in decision making was notably higher in the treatment group compared to the control at 24 hours after the surgery. This is expected as we have encouraged patients in the treatment group not to be passive and shy, rather to participate actively in the choice and manner of pain management. The goal of encouraging patients to participate in decision-making is to increase satisfaction and better health outcomes. Studies have also hinted this even can reduce the patient report of pain intensity [27,54] and well randomized controlled trials are also currently investigating the topic [55].
Our results from the mediation analysis, however, revealed insignificant indirect effect, for both pain intensity and patient satisfaction, and patient participation in decision did not mediate the treatment with both outcome measures. Still, our result should not be over-emphasized. The lack of mediating effect could also be due to unmeasured confounder between mediator and outcome and this is very difficult to rule out. Also, the absence of statistically significant mediating effects identified could be due to the study being underpowered to detect these effects, as the mediation analysis was secondary and was not powered for this analysis. However, we have measured the most important predictors of severe postoperative pain as identified from systematic review except for preoperative anxiety level. These also were appropriately tested if the addition of such measured confounder covariates-(age, chronic pain, types of surgery, types of anesthesia and duration of surgery) -affected the mediation and the results were the same. A previous study also showed that higher patient-driven participation in decision-making was associated with lower odds (OR, 0.82; 95% CI, 0.75-0.89) of frequent pain, but was not significantly associated with severity of pain. Interestingly they have found no significant association with either frequency or severity of pain when the patient participation was physician-driven [56]. limitations that are worthy to note. There was a clear a baseline imbalance between the control and treatment groups, which could bias results to a larger extent and caution is mandatory when interpreting results. Even though we have dealt this during the treatment effect estimation, it is still of a concern for the internal validity of the study. Another important limitation of this study is that the association between preoperative patient education and pain intensity or patient education and patient satisfaction, the mechanisms underlying this association may be complicated. Future research should consider a number of other potential mediators of the above-mentioned association. Even patient participation in decision making should be tested with a more powerful study designs [57]. Heterogeneous sample from different surgical categories might also affect the results in the same manner. This has been also raised previously as a concern from previous trials dealing the same topic [58]; but it could contribute positively to external validity and generalizability of the study. Aside from this, there are known threats to internal validity when one is implementing a quasi-experiment study design. We have tried to control for most threats using various methods. In this regard, the use of two control groups adequately controlled for what called "history effect" [59]. Maturation also seems not to affect the trial as the duration of the study was short [11,60]. Patients are the only one who was blinded so there is a threat of Hawthorne effect [59].

Conclusions
To sum up, this study has tested the impact of the educational intervention, in decreasing postoperative pain intensity by encouraging patients to be part of the solution (participate in the care). The treatment was successful in increasing patients participation in decision making, as anticipated. However, its impact on decreasing pain intensity was only noted at the last measurement point after surgery. Patient participation in decision making not mediated the treatment with pain intensity, even though demonstrated the same pattern. Helsinki. A comprehensive oral and written patient information of the purpose and procedure of the study was given. Before inclusion of participants, informed consent was obtained, and patients were informed about their rights to refuse or withdraw at any given time. Confidentiality of the individual information gathered was discussed, and additionally, any personal information was anonymized before the final analysis.

Consent for publication
Not Applicable Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing Interests
The authors declare that they have no competing interests    Within-subjects mediation for satisfaction (see Bolger and Laurenceau, 2013): To reduce confusion, we have omitted time as a predictor and we treat X, M, and Y as varying within-subjects only. *Adjusted for age, sex, preoperative pain, type of surgery, type of anesthesia, baseline worst pain intensity and duration of surgery.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.