Study design

This study employed data of a large prospective registry cohort called DATAPAIN, which contains patient reported outcome measures (PROMs), socio-demographic factors and chronic pain-related data.⁷ The cohort entails patients with chronic pain referred to a Pain Centre since 2003 and are requested to complete the DATAPAIN questionnaire before their first consultation. As of March 2016, treatment outcomes of the patients with chronic pain were also gathered. For this article, data collected between March 2016 and June 2020 were used in the longitudinal analysis, with data before consultation (baseline) and 6 months later (follow-up). Ethical clearance was obtained from the medical ethical committee (METC 2020-1568).

Study population

Patients suffering from chronic pain who are referred to the Pain Centre are requested to complete a pain questionnaire before their first consultation. Collecting information conforms the biopsychosocial model on pain characteristics, quality of life, and psychological aspects. Additionally, demographic characteristics, such as sex, age, level of education, marital status and employment, were gathered. At 6 months follow-up, patients are asked to fill out a second pain questionnaire, with questions concerning the received pain care.¹² For this study, 1637 patients gave additional informed consent to be contacted at 6 months follow-up, had an age above 18 years and had pain at baseline for more than 3 months. This population was subdivided into complex and non-complex cases based on criteria discussed in Complex versus non-complex cases section.

DATAPAIN

The DATAPAIN registry cohort is part of a quality registration, and the results of the questionnaires are used in the intake (screening and diagnosis) of patients suffering from chronic pain. The set of questionnaires collected at baseline in the DATAPAIN registry is compiled in accordance with the biopsychosocial model. These are all validated questionnaires for the chronic pain population and suggested by the IMMPACT recommendations.¹¹ To allow proper quality improvement, the registry was expanded with outcome measurements and quality indicators from a patient's perspective. This registry is followed nationally, and since 2020 collecting patient-reported outcome measurements (before consultation and 6 months after treatment) is mandatory for all Dutch registered pain centers (Quality Registry Pain Medicine).¹³

We currently use this set in both research and clinical settings: diagnostic (screening) phase, multidisciplinary consultation, experience, and outcome of treatment. In our practice, patients with chronic pain are treated within a multidisciplinary team with a biopsychosocial approach, providing tailored personalized care. To be precise, at intake, the pain specialist evaluates the results of the baseline questionnaire together with the patient. The results are considered in the anamnesis and treatment protocol that is elaborated in a shared decision with the patient.

Measurements

Complex versus non-complex cases

Based on clinical expertise the complex cases were defined, as those who presented an average pain intensity score of 7–10 (≥ 7) on the NRS, in combination with a depression or anxiety score above 10 (> 10) on the Hospital Anxiety and Depression Scale (HADS), and in combination with pain catastrophizing of above 31 (> 31) on the Pain Catastrophizing Scale (PCS). These cut-off scores are based on the recommendations of Boonstra et al., Stern at al., and Sullivan et al.^16-19

Statistical analysis

The data were tested on normal distribution for the statistical analysis, by histograms, scatterplots, and QQ-plots. The above-mentioned variables were described with mean and standard deviations, count, and proportion for the continuous and binary variables, respectively. Subsequently, analyses of differences between the complex and non-complex cases were performed with the independent sample t-test or chi-square test depending on whether the variables were continuous or binary of kind.

Logistic regression was applied to determine the likelihood of success for both complex and non-complex cases, by calculating odds ratios (OR) with confidence intervals (CI) for the PGIC and the GPH. Moreover, the change scores between baseline and follow-up of the NRS and BPI were calculated and applied as the dependent variables of the linear regression. To observe, if both complex and non-complex cases have the same possibility of the specified difference score.

In both logistic and linear regression analyses, a backward stepwise elimination was conducted to select potential confounding variables. The confounders considered were those presented in Table 1: the demographic variables and pain characteristics. These factors are associated with (the maintenance of) chronic pain.²⁰ The results of the statistical analyses were evaluated on clinical relevance and preciseness of measurement, recommended by the American Statistical Association.²¹ The statistical analyses were executed in R, a language for statistical computing, through R version 3.6.1 with R Studio1.4.

Description of the population

The summary table (Table 1) displays the socio-demographics, pain characteristics and measurements of the study population, consisting of 1637 patients with chronic pain who had completed both pre- and post-treatment questionnaires. Figure 1 shows a flowchart of the chronic pain patients included in both complex and non-complex cases. The complex cases were represented by 345 patients (21%), with an average age of 61 years and consisted of 50% of men. Twenty-nine percent of the complex cases had no relationship, 89% had a low level of education and 77% had no paid job. The non-complex cases consisted of 1291 patients (79%), with an average age of 60 years and of which 41% were male. The sociocultural variables statistically differed with the complex cases; 22% had no relationship status, 73% had a low level of education⁷ and 67% had no paid job at intake. Pain characteristics of both groups are shown in Table 1.

Primary outcomes: pain relief on the NRS and treatment success on the PGIC

Differences in outcome of the PGIC for complex and non-complex group

At follow up, the general success rates of the PGIC were significantly lower for the complex cases (p < 0.001). Figure 2 is a visual presentation of the outcome of this patient-rated measure, in which can be observed that the complex cases were distributed more among the ineffective categories of the PGIC. This observation was confirmed by the odds ratio, resulting in an adjusted OR on the PGIC of β = 0.52 (95% Confidence Interval [CI]: 0.37–0.78) for the complex cases (Table 3).

TABLE 3. Results of the logistic regression analyses of the PGIC.

PGIC	B (SE)	p-Value	95% CI for odds ratio (OR)
			Lower bound	OR	Upper bound
Crude model Complex cases	−0.655 (0.19)	< 0.001	0.37	0.52	0.74
Adjusted model Complex cases	−0.646 (0.18)	< 0.001	0.37	0.52	0.78

• Note: Results of the logistic regression of PGIC for a crude and adjusted model. The latter was adjusted for the variables that were statistically significant in the backward stepwise elimination, the variables were as follows: sex(women)*, pain duration > 1 year***. *p-value < 0.05, **p-value < 0.01, and ***p-value < 0.001.
• Abbreviations: B, beta coefficient; OR, odds ratio; PGIC, patient global impression to change; SE, standard error.

Secondary outcomes: general health on the GPH and pain interference in daily life on the BPI

Differences in outcome of the GPH for complex and non-complex group

The positive general health of the GPH was significantly lower for the complex cases (p < 0.001). Figure 3 visualizes that the complex cases are distributed more among the ineffective categories of the GPH. Confirmed by the adjusted OR on the GPH of β = 0.32 (95% CI: 0.15–0.61) (Table 4).

TABLE 4. Results of the logistic regression analyses of the GPH.

GPH	B (SE)	p-Value	95% CI for odds ratio (OR)
Crude model			Lower bound	OR	Upper bound
Crude model Complex CP	−1.275 (0.37)	< 0.001	0.12	0.28	0.55
Adjusted model Complex CP	−1.134 (0.35)	0.001	0.15	0.32	0.61

• Note: Results of the logistic regression of GPH for a crude and adjusted model. The latter was adjusted for the variables that were statistically significant in the backward stepwise elimination, the variables were: work(having paid work)*, pain duration > 1 year***. *p-value < 0.05, **p-value < 0.01, and ***p-value < 0.001.
• Abbreviations: B, beta coefficient; GPH, general perceived health; OR, odds ratio; SE, standard error.

Differences and similarities of the BPI for the complex and non-complex group

At baseline, the BPI-WAW showed an average difference between groups of almost two points with a mean of 5.69 (SD: 2.38) for the non-complex cases and 7.58 (SD: 1.76) for the complex cases (p < 0.001; Table 1). This difference remained at follow-up, as the change in score between baseline and follow-up on the BPI-WAW was similar (non-complex: −0.77 [SD: 2.54] and complex: −0.93 [SD: 2.1] with a p-value of 0.231). Furthermore, the linear regression of the BPI-WAW indicated no statistically significant association with an adjusted coefficient of −0.21 (SE: 0.15); p: 0.153, (Table 5).

TABLE 5. Results of linear regression analyses for the BPI change on both subscales.

BPI difference scoring	Baseline – follow up	B (SE)	p-Value
BPI WAW change	Crude model Complex cases	−0.162 (0.15)	0.281
	Adjusted model Complex cases	−0.213 (0.15)	0.153
BPI REM change	Crude model Complex cases	−0.556 (0.163)	< 0.001
	Adjusted model Complex cases	−0.570 (0.145)	0.002

• Note: Results of the linear regression of BPI subscales for a crude and adjusted model. The latter was adjusted for the variables that were statistically significant in the backward stepwise elimination, the variables included in the BPI WAW were: pain duration > 1 year***, pain location chest**. For the BPI REM the variables adjusted for were: pain duration > 1 year***. *p-value < 0.05, **p-value < 0.01, and ***p-value < 0.001.
• Abbreviations: B, Beta coefficient; BPI, Brief pain inventory; REM, relations with others, enjoyment of life, mood and sleep; SE, Standard error; WAW, walk, general activity, work.

The mean BPI-REM score at baseline was 3.69 (SD: 2.27) for the non-complex cases and 6.72 (SD: 1.97) for the complex cases (Table 1). The difference score of the complex cases (−0.98 [SD: 2.44]) was twice as large as for the non-complex cases (−0.43 [SD: 2.38]) (p: 0.004). But the difference between groups remained existent at follow-up (Figure 4). Linear regression confirmed the findings of the BPI difference scores: the adjusted regression coefficient for the complex cases was −0.57 (SE: 0.15); p: 0.002 on the BPI-REM.

Primary outcomes

Secondary outcomes

Limitations

The most challenging limitation of this study is the fact that other important factors were not routinely collected, due to the limitations of a registry cohort. The analyses could have profited from confounders such as the classification of disease (ICD), intervention and current medication intake. Besides, we can assume that the complex cases did receive more psychological and psychiatric care, that may lead to longer time before results are visible. However, we have adjusted our findings for many potential confounding variables and baseline differences in the multivariable analyses. The study is a large academic prospective cohort with a heterogeneous population in which multiple PROMs cover many aspects of chronic pain, making cross-sectional and longitudinal analyses possible.

Future perspective

There is need for further clinical research to observe if similar findings exist in the likelihood of treatment outcomes for this group. This information may contribute to the search for relevant predictors and/or treatments of which the (complex) patients with chronic pain may benefit as for example in improving treatment satisfaction and quality of life. The similarities in reduction of pain relief and physical interference between the complex and non-complex cases may imply that the multidisciplinary care provided is adequate. However, the care provided may not be sufficient for improving treatment satisfaction and quality of life. The focus of multidisciplinary tertiary pain clinics is on pain relief, physical, psychological improvement, yet topics that may improve quality of life like societal participation are lacking. A suggestion would be to combine the personalized multidisciplinary approach together with a rehabilitation and societal approach. Social workers may provide help on topics such as education and work. Even more reason to do so is because, besides the cognitive-affective risk factors and pain severity, a high number of patients with low levels of education and absence of employment are present in the complex cases. This may hinder the usage of the health care system and full understanding of the information provided.^38-41 Hence, interventions, such as pain education, self-management, and specific information for family and caregiver(s), may improve patient outcomes.

In conclusion, this study indicates that psychological health status in combination with pain severity, reduces the likelihood on treatment satisfaction, general perceived health, and pain interference of emotional activities at six months follow up. Yet, belonging to the complex cases does not interfere with the change in score of pain relief nor of pain interference of the physical activities. When treating complex patients, the desired treatment outcome(s) should be recognized by the specialist and patient before initiating treatment, as these may be less likely to occur depending on the outcome and may improve the quality of pain management, such as in monitoring pain processes and reassessing pain management plans to the needs of the patients.