Defining chronic pain by outcome probability

A prognostic approach to defining chronic pain combines information on past and current pain status, with information on other prognostic variables, into a definition based on outcome probabilities. This approach was initially developed by Von Korff and Miglioretti using data from a cohort of primary care back pain consulters in Washington State, USA (Von Korff and Miglioretti 2005, 2006), and was subsequently replicated across several pain conditions in independent studies (Dunn et al. 2008; Thomas et al. 2008; Von Korff and Dunn 2008).

Development of the approach

The definition of chronic pain used in the development of the approach was clinically significant pain likely to be present 1 or more years in the future. This outcome (clinically significant pain) was operationalized using Chronic Pain Grades II to IV (Von Korff et al. 1992) at the follow-up time points. This identifies people with moderate to severe pain intensity and mild to severe dysfunction 1, 2, or 5 years after the baseline evaluation, which typically occurred about 2-4 weeks after an index pain visit in a primary care setting.

In the initial work, latent transition regression analysis was used to identify categories of latent pain severity: no pain, mild pain, moderate pain/limitation, and severe limiting pain. The observed probabilities of severe pain at the subsequent observation demonstrated the utility of pain severity measures in predicting risk of future clinically significant pain, as people with severe pain tended to continue to have severe pain, and people with mild or moderate pain were very unlikely to develop severe pain. This observation suggested that chronicity could be measured by the probability of clinically significant pain continuing at a future point in time.

The next step was to consider prognostic variables other than pain severity, to further differentiate people with lower or higher risk of future clinically significant pain. The variables considered were: depression, measured using the SCL-90-R (Derogatis et al. 1974), the number of days with pain in the previous 6 months, and the number of other pain sites (headache, abdominal pain, chest pain, and facial pain). These variables were then combined with ratings of pain intensity and interference with activities to produce a risk score. Table 2.1 shows the scoring rules used to calculate the risk score for each patient. The items were scored using a limited number of categories established by examining distributional properties of individual items, with 0 scores assigned to low values. It was subsequently assessed whether the predictive accuracy of scoring could be improved by optimizing the scoring weights for each of the items. It was found that predictive validity was not materially improved by using regression-based scoring weights.

The ability of the risk score to predict chronic pain at follow-up was then assessed using a smoothed plot of the baseline risk score against the probability of clinically significant pain 1 year after the index consultation. Probable chronic pain was defined by an 80% or greater probability of future clinically significant pain, possible chronic pain as a 50% or greater probability, and

Table 2.1 Scoring rules for estimating prognostic risk score

Item

Item value

Risk score value

Average pain intensity

  • 0-3
  • 4-6
  • 7-10
  • 0
  • 1
  • 2

Worst pain intensity

  • 0-4
  • 5-7
  • 8-10
  • 0
  • 1
  • 2

Current pain intensity

  • 0-2
  • 3-4
  • 5-10
  • 0
  • 1
  • 2

Interference with usual activities

  • 0-2
  • 3-4
  • 5-10
  • 0
  • 1
  • 2

Interference with work/household activities

  • 0-2
  • 3-4
  • 5-10
  • 0
  • 1
  • 2

Interference with family/social activities

  • 0-2
  • 3-4
  • 5-10
  • 0
  • 1
  • 2

Days of activity limitation due to pain in prior 3 months

  • 0-2
  • 3-6
  • 7-15
  • 16-24
  • 25-90
  • 0
  • 1
  • 2

3

4

SCL-90-R+ Depression score

<0.50 0.50-<1.0

  • 1.0— <1.5 1.5—<2.0
  • 2.0- 4.0
  • 0
  • 1
  • 2

3

4

Number of other pain sites

  • 0
  • 1
  • 2

3

4

  • 0
  • 1
  • 2

3

4

Table 2.1 (continued) Scoring rules for estimating prognostic risk score (Continued)

Item

1 tem value

Risk score value

Number of days with index pain in prior 6 months

0-30

0

31-89

1

90-120

2

121-160

3

161-180

4

Total risk score

0-28

  • * Risk score values are summed across items. ф Measured on 0-10 numerical rating scale.
  • 3 Symptom Checklist-90-R (Derogatis et al. 1974).

intermediate risk as a 20% or greater probability of future clinically significant pain, with low risk below 20%. Cut-points on the risk score were then determined from the probability plot to define chronic pain status in terms of the probability of having clinically significant pain at a future time point. Low risk was determined as risk scores of 0-7 (31% of the sample), intermediate risk from 8-15 (43%), possible chronic pain as a score of 16-21 (20%) and probable chronic pain as a risk score of 22 or more (6% of the sample). These categories strongly predicted clinically significant pain at 1 year, as would be expected (see Table 2.2), but they also strongly predicted clinically significant pain at the 2-year and 5-year follow-ups.

These findings provided initial support for the predictive validity of a prognostic approach to defining chronic pain, and supported a multivariable approach to defining chronic pain in terms of outcome probabilities.

 
Source
< Prev   CONTENTS   Source   Next >