Table 4.

Predictors of all-cause direct costs after adjusting for baseline characteristics*. Direct costs were calculated as the sum of all-cause medical and pharmacy costs. Outcomes were predicted using a generalized linear model with γ distribution and log link function.

PredictorAS, n = 4078, Coefficient Estimate (95% CI)**pPsA, n = 10,277, Coefficient Estimate (95% CI)**p
Age, yrs, reference ≥ 65 yrs
  18–341.28 (1.09–1.50)< 0.011.18 (1.07–1.31)< 0.01
  35–441.18 (1.02–1.36)< 0.051.07 (0.98–1.16)NS
  45–541.20 (1.05–1.38)< 0.011.09 (1.01–1.18)< 0.05
  55–641.30 (1.13–1.50)< 0.0011.14 (1.06–1.22)< 0.001
Sex, reference = female
  Male1.02 (0.94–1.10)NS1.03 (0.99–1.07)NS
Region in the US, reference = West
  Northeast0.98 (0.86–1.11)NS1.03 (0.96–1.11)NS
  North Central1.01 (0.90–1.12)NS0.88 (0.83–0.94)< 0.001
  South0.88 (0.80–0.97)< 0.050.93 (0.87–0.98)< 0.05
Insurance plan*, reference = HMO and POS capitation
  FFS0.91 (0.82–1.10)≥ 0.051.04 (0.98–1.10)NS
Baseline ECI score, reference = 0
  10.99 (0.87–1.13)NS1.09 (1.04–1.15)< 0.01
  21.25 (1.09–1.44)< 0.011.21 (1.13–1.28)< 0.001
  31.26 (1.07–1.48)< 0.011.37 (1.25–1.49)< 0.001
  42.12 (1.81–2.48)< 0.0011.95 (1.78–2.13)< 0.001
Baseline biologics use, reference = patients with no biologic use2.82 (2.61–3.05)< 0.0013.12 (2.98–3.25)< 0.001
  • * Patients with missing region (n = 42 for AS, < 1%; n = 125 for PsA, < 1%), missing insurance type (n = 157 for AS, < 3%; n = 425 for PsA, < 4%), or zero outcome (n = 11 for AS, < 1%; n = 5 for PsA, < 1%) were excluded from the multivariate analysis.

  • ** Coefficient estimates were the exponentials of the original estimates. A coefficient of > 1.0 indicates higher cost and < 1.0 indicates lower costs for a given variable compared with the specified “reference.” AS: ankylosing spondylitis; PsA: psoriatic arthritis; FFS: fee for service; ECI: Elixhauser Comorbidity Index; NS: not significant; HMO: health maintenance organization; POS: point of service.