Multivariate correlations between pain, life interference, health-related quality of life and full-time sick leave 1 year after multimodal rehabilitation, focus on gender and age

Figures & data

Figure 1. Drop outs.

Table 1. Variables and instruments in this study.

Sick leave	In Sweden, a person can be on 100% sick leave or partial sick leave (25%, 50%, 75 %) based on the actual level of employment and work ability. The variable before MMRP was scaled as no sick leave = 0 and full-time sick leave = 4. The Y-variable, degree of sick leave one year after MMRP, was dichotomised into either no full-time sick leave = 0 or full-time sick leave = 1
Employed	The variable is scaled as unemployed = 0 and full-time employed = 1. In Sweden, it is possible to be on part-time or full-time sick leave when you are unemployed
Work importance	The participants were asked: “What significance does work have for you in addition to being importance as a source of income?” Answers were ranked in a scale from 1 to 5 (1 = “Very important” and 5 = “Not at all important”)
Belief in recovery	Belief in recovery, was assessed with the question: How convinced are you about recovery? On a scale from 1-5 (1= “Entirely convinced about recovery”; 5= “Not all convinced about recovery”)
Pain duration	Pain duration in days
Number of pain sites	The number of pain sites was reported using 36 predefined anatomical areas: 18 on the left side of the body and 18 on the right side of the body. The sum of the reported anatomical areas was used in this study
The Numeric Pain Rating Scale (NPRS)	Pain intensity for the last 7 days, 11-point scale 0-10. A high value indicates a higher level of pain [4]
EuroQol 5 D (EQ5D)	Identifies patients’ perceived state of health in two parts. The first part defines five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Dimensions are converted to an index (EQ5D index) ranging from −0.594 to 1 (1= indicates full health, 0 indicates dead, below 0 indicates worse than dead) EQ-5D VAS is a 100-point scale, with defined end points. High values indicate good health while low values indicate poor health [30]
Hospital Anxiety and Depression Scale (HADS)	Used to detect symptoms of depression and anxiety. Fourteen items divided into two subscales. The range for each subscale is 0–21. Score ≥ 11 indicates a definite case in each scale [31]
Multidimensional Pain Inventory (MPI)	Measures psychosocial, cognitive and behavioural effects of chronic pain. Sixty-one items divided into three groups: one psychosocial and two behavioural parts. Seven-pointed numeric scale, range 0–6. A higher score indicates higher pain intensity and a higher impact on interference in daily life or higher affective distress. The scale Life control scale is the opposite: a higher score indicates better control in life. Five variables were used in this study: (1) Pain severity: assesses perceived severity of pain; (2) Life interference: assesses the individual’s perceptions regarding how pain interferes with their lives including interference with family, work and social recreational activities and satisfaction; (3) Life control: assesses the perceived life control, incorporating the perceived ability to solve problems and feelings of mastery and competence; (4) Affective distress: ratings of depressed mood, irritability and tension; (5) Social support: appraisal of support received from spouses, family and significant others, such as being concerned supportive and attentive [32]
The Short Form health survey (SF-36)	Self-rated health status that is intended to represent a person’s conditions and limitations in daily life on eight scales. (1) Physical functioning; assesses the perceived performance of physical activities including bathing or dressing. (2) Role-physical: appraises the performance of work or other daily activities as a result of physical health; (3) Bodily pain: assesses limitations due to pain; (4) General health: determines the perceived health and belief that the state of health will probably get worse; (5) Vitality: assesses the feeling of being tired and worn out; (6) Social functioning: appraises the interference with normal social activities because of physical and emotional problems; (7) Role-emotional: assesses perceived performance with work or other daily activities as a result of emotional problems; and (8) Mental health: assesses feelings of nervousness and depression. A higher score indicates better health status (0–100) [33]

The instruments have been used in their entirety, unless stated otherwise in the table.

Figure 2. The plot showed a validate plot (after 100 permutations) of the Orthogonal Partial Least Square regression model (OPLS) for the selected Y (full-time sick leave) for: (a) all participants; (b) women; (c) men; (d) age group 31–45 years and (e) age group 46–65 years. The horizontal axis showed the correlation between the permuted Y vectors and the original vector of the full-time sick leave. The original Y had the correlation 1.0 with itself and defined the high point of the horizontal axis. For validity, the values of R2 and Q2 for the original model on the right are higher than the permuted model to the left or the Q2 regression line intersects the vertical axis at or below zero.

Table 2. Participants’ characteristics.

	Total	Women	Men	p Value	18–30 years	31–45 years	46–65 years	p Value
	n = 284	n = 237	n = 47		n = 36	n = 103	n = 145
Age (years mean (SD), n = 284	45 (10.9)	46 (11.1)	46 (9.8)	0.451^a	25 (2.8)	39 (4.5)	53 (4.8)	-
Born in Nordic countries, n (%) n = 284	277 (97)	231 (98)	46 (98)	0.870^b	35 (97)	100 (97)	142 (98)	0.907^b
University education, n (%) n = 283	101 (36)	95 (40)	6 (13)	<0.001^b	13 (37)	41 (40)	47 (32)	0.479^b
Employed, n (%) n = 284	190 (67)	164 (69)	26 (55)	0.065^b	13 (36)	75 (73)	102 (70)	<0.001^b
Consulted a doctor more than four times (%) n = 280	199 (70)	171 (73)	28 (60)	0.057^b	27 (77)	72 (71)	100 (70)	0.694^b
Work importance^1, n (%) n = 279	234 (82)	196 (84)	38 (84)	0.909^b	30 (83)	94 (91)	110 (79)	0.029^b
Belief in recovery^2 n (%) n = 266	41(14)	36 (16)	5(12)	0.493^b	4 (11)	18 (17)	19 (13)	0.517^b
Number of pain sites, mean (SD) n = 284	16 (8)	16(7.7)	13(7)	0.043^a	15 (8.2)	16 (8.2)	15 (7.2)	<0.001^d
Duration of pain, month, mean (SD) n = 230	91.5 (88.6)	84.4 (82)	122 (109)	0.035^a	29 (22)	45.1 (85.2)	64.1 (95.5)	<0.001^d
Numeric Pain Rating Scale (NPRS)
Pain intensity last week, mean (SD) 283	6.86 (2)	7 (2)	7 (3)	0.673^b	7 (2)	7 (2)	7 (2)	0.899^e
EuroQol 5 D-(EQ5D)
EQ-5D index, median (IQR) n = 282	0.157 (0.532)	0.159 (0.532)	0.099 (0.092)	0.006^b	0.157 (0.205)	0.157 (0.408)	0.159 (0.316)	0.465^e
EQ-5D VAS, median (IQR) n = 277	40 (27)	40 (25)	32.5(26.75)	0.142^b	40 (25)	35 (20)	40 (13.75)	0.570^e
Hospital Anxiety and Depression Scale (HADS)
Anxiety, mean (SD) n = 282	7(7)	7(7)	6 (6)	0.430^b	8.5 (6.5)	8 (5)	6 (7)	0.019^ef
Depression, mean (SD) n = 283	7 (5)	7 (5)	7 (8)	0.225^b	7.5 (6)	8 (5)	7 (6)	0.351^e
Multidimensional Pain Inventory (MPI)
Life interference, median (IQR) n = 281	4.64 (1.19)	4.55 (1.18)	4.9 (0.97)	0.037^b	4.45 (1.18)	4.73 (1.09)	4.55 (1.3)	0.459^e
Pain severity, median (IQR) n = 284	4.33 (1)	4.33 (1)	4.33 (1.33	0.681^b	4.5 (1.33)	4.33 (1)	4.33 (1)	0.896^e
Affective distress, median (IQR) n = 284	3.33 (1.67)	3.33 (1.67)	3 (1.67)	0.709^b	3.33 (2)	3.33 (1.33)	3 (2)	0.045^ef
Life control, median (IQR) n = 283	3 (1.25)	3 (1.25)	2.75 (1.5)	0.955^b	2.75 (1.18)	2.75 (1.18)	3 (1)	0.092^e
Social support, median (IQR) n = 282	4.33 (2)	4(2)	4.67 (1.33)	0.121^b	4.33 (1.59)	4 (1.83)	4.33 (2.33)	0.678^e
Short Form Health Survey (SF-36)
Physical functioning, median (IQR) n = 283	55 (25)	55(15)	55 (30)	0.755^b	61.5 (33.75)	55 (30)	50 (25)	0.002^ef
Role physical, median (IQR) n = 282	0 (25)	0 (25)	0 (25)	0.872^b	0 (25)	0 (0)	0 (25)	0.668^e
Bodily pain, median (IQR) n = 283	22 (11)	22 (11)	22(19)	0.115^b	22 (20)	22 (11)	22 (11)	0.821^e
General health, median (IQR) n = 282	40 (25)	40 (25)	35(20)	0.501^b	40 (29.25)	40 (20)	40 (25)	0.855^e
Vitality, median (IQR) n = 283	20 (23)	20 (20)	25 (25)	0.378^b	25 (28.75)	15 (20)	22.5 (26)	0.077^e
Social functioning, median (IQR) n = 283	50 (38)	50 (38)	50(50)	0.528 ^f	50 (38)	50 (38)	50 (46.75)	0.205^e
Role emotional, median (IQR) n = 280	67 (100)	67 (28)	67(100	0.907^b	67 (100)	67 (100)	67 (100)	0.748^e
Mental health, median (IQR) n = 283	64 (24)	64 (28)	68(20)	0.967^b	62 (27)	60 (20)	68 (24)	0.001^ef

Categorical variables presented with n %. Scale variables presented with mean and SD. Ordinal variables presented with median interquartile range q1–q3. ^at-test^{; b}X²; ^cMann–Whitney;^d One-Way ANOVA; ^eKruskal–Wallis Test: ^fBonferroni correction for multiple. ¹Included answer: very important (1) and extremely (2), ²included answer: totally convinced (1) and convinced (2). Significant values are shown in bold (p = 0.05); EQ5D: EuroQol 5D; HAD: Hospital Anxiety and Depression Scale; MPI: Multidimensional Pain Inventory; NRPS: Numerical Rating Pain Scale; SF-36: Short Form Health Survey.

Table 3. Sick leave before and one year after MMRP .

Sick leave before MMRP	Total	Women	Men		18–30 years	31–45 years	46–65 years
	n = 284	n = 237	n = 47	p Value	n = 36	n = 103	n = 145	p Value
No sick leave	100 (35)	78 (33)	22 (47)	0.068	20 (56)	35 (34)	45 (31)	0.021
25%, n (%)	17 (6)	17 (7)	0 (0)	0.042	1 (3)	7 (7)	9 (6)	0.673
50%, n (%)	43 (15)	39 (17)	4 (9)	0.165	2 (6)	17 (17)	24 (17)	0.229
75%, n (%)	21 (7)	17 (7)	4 (9)	0.749	4 (11)	8 (8)	9 (6)	0.593
Full-time sick leave	103 (36)	86 (36)	17 (36)	0.988	9 (25)	36 (35)	58 (40)	0.231
Sick leave 1 year after MMRP
Full-time sick leave	116 (41)	94 (40)	22 (47)	0.363	13 (36)	40 (39)	63 (43)	0.635

Significant values are shown in bold (p = 0.05).

Figure 3. Combined OPLS loading column plot for the predictive component in the model for full-time sick leave 1 year after MMRP with 95% jack-knife confident interval bars. Relationship between the X variables (loadings before MMRP, light grey) and the loading for the Y variable (full-time sick leave 1 year after MMRP, dark grey). Only significant variables are shown in the model. In the figure, column height indicates the importance of each variable. Values closer to zero being less important irrespective of sign (±). Same sign for Y variable and X variable means they are positively correlated. High loadings of the Y variable indicate high correlation with the predictive component and X. Model variance R²Y =26% and prediction value Q = 20%.

Figure 4. Combined OPLS loading column plot for the predictive component in the model for full-time sick leave 1 year after MMRP with 95% jack knife confident interval bars for women. Relationship between the X variables (loadings before MMRP, light grey) and the loading for the Y variable (full-time sick leave 1 year after MMRP, dark grey). Only significant variables are shown in the model. In the figure, column height indicates the importance of each variable. Values closer to zero being less important irrespective of sign (±). Same sign for Y variable and X variable means they are positively correlated. High loadings of the Y variable indicate high correlation with the predictive component and X. Model variance R²Y = 26.3% and prediction value Q = 13.5%.

Figure 5. Combined OPLS loading column plot for the predictive component in the model for full-time sick leave 1 year after MMRP with 95% jack knife confident interval bars for men. Relationship between the X variables (loadings before MMRP, light grey) and the loading for the Y variable (full-time sick leave 1 year after MMRP, dark grey). Only significant variables are shown in the model. In the figure, column height indicates the importance of each variable. Values closer to zero being less important irrespective of sign (±). Same sign for Y variable and X variable means they are positively correlated. High loadings of the Y variable indicate high correlation with the predictive component and X. Model variance R²Y = 28% and prediction value Q = 2%.

Figure 6. Combined OPLS loading column plot for the predictive component in the model for full-time sick leave 1 year after MMRP with 95% jack-knife confident interval bars for age group 31–45 years. Relationship between the X variables (loadings before MMRP, (light grey) and the loading for the Y variable (full-time sick leave 1 year after MMRP, dark grey). Only significant variables are shown in the model. In the figure, column height indicates the importance of each variable. Values closer to zero being less important irrespective of sign (±). Same sign for Y variable and X variable means they are positively correlated. High loadings of the Y variable indicate high correlation with the predictive component and X. Model variance R²Y = 18.3% and prediction value Q = 10%.

Figure 7. Combined OPLS loading column plot for the predictive component in the model for full-time sick leave 1 year after MMRP with 95% jack-knife confident interval bars for the age group 46–65 years. Relationship between the X variables [loadings before MMRP, (light grey) and the loading for the Y variable (full-time sick leave 1 year after MMRP, dark grey)]. Only significant variables are shown in the model. In the figure, column height indicates the importance of each variable. Values closer to zero being less important irrespective of sign (±). Same sign for Y variable and X variable means they are positively correlated. High loadings of the Y variable indicate high correlation with the predictive component and X. Model variance R²Y = 37% and prediction value Q = 22%.

Breivik H, Collett B, Ventafridda V, et al. Survey of chronic pain in Europe: prevalence, impact on daily life, and treatment. Eur J Pain. 2006;10(4):287–333.

 PubMed Web of Science ®Google Scholar

van Agt HM, Essink-Bot M-L, Krabbe PF, et al. Test-retest reliability of health state valuations collected with the EuroQol questionnaire. Soc Sci Med. 1994;39(11):1537–1544.

 PubMed Web of Science ®Google Scholar

Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand. 1983;67(6):361–370.

 PubMed Web of Science ®Google Scholar

Kerns RD, Turk DC, Rudy TE. The west haven-yale multidimensional pain inventory (WHYMPI). Pain. 1985;23(4):345–356.

 PubMed Web of Science ®Google Scholar

Sullivan M, Karlsson J, Ware JJ. The Swedish SF-36 Health Survey – I. Evaluation of data quality, scaling assumptions, reliability and construct validity across general populations in Sweden. Soc Sci Med. 1995;41(10):1349–1358.

 PubMed Web of Science ®Google Scholar