Skip to main content
. 2020 Sep 18;20:883. doi: 10.1186/s12913-020-05723-3

Table 3.

Marginal effects estimated by GEE regression of DOT/1000 resident days unadjusted and adjusted for resident characteristics and comorbidities

Unadjusted model Adjusted model
Marginal effect 99% CI p-value Marginal effect 99% CI p-value
Year 4.09 1.18, 6.99 < 0.001 3.12 −0.05, 6.29 0.011
Month
 January Reference Reference
 February −0.19 −3.71, 3.33 0.889 0.98 −3.01, 4.96 0.527
 March −0.67 −4.71, 3.38 0.672 −0.03 −4.53, 4.46 0.985
 April −2.08 −6.12, 1.96 0.185 −1.23 −5.52, 3.06 0.459
 May 3.52 −1.06, 8.11 0.048 5.14 0.2, 10.08 0.007
 June 3.95 −0.78, 8.69 0.032 5.51 0.32, 10.7 0.006
 July 8.27 3.32, 13.21 0.000 9.70 4.37, 15.03 < 0.001
 August 10.51 5.38, 15.63 0.000 12.36 6.91, 17.81 < 0.001
 September 5.48 0.79, 10.17 0.003 7.47 2.36, 12.59 < 0.001
 October 0.01 −4.65, 4.66 0.998 1.36 −3.79, 6.51 0.496
 November 0.06 −4.47, 4.6 0.971 1.83 −3.25, 6.92 0.352
 December 3.38 −0.78, 7.55 0.036 5.40 0.59, 10.2 0.004
WHO Watch List Antibiotic 219.56 187.14, 251.97 < 0.001 247.04 206.61, 287.47 < 0.001
WHO Watch List antibiotic x year −0.41 −5.96, 5.14 0.849 −1.64 −7.96, 4.67 0.503
Age 2.28 0.00, 3.91 0.000
Age at admission −0.54 −2.05, 0.97 0.358
Men −2.48 −12.99, 8.02 0.542
Comorbidities
 Dementia −9.93 −19.12, −0.74 0.005
 Chronic respiratory disease 27.48 16.66, 38.31 < 0.001
 Urinary incontinence 3.36 −4.57, 11.28 0.275
 History of UTI 50.18 32.01, 68.36 < 0.001
 History of wound 28.78 1.28, 56.28 0.007
 History of SSTI 30.07 5.28, 54.86 0.002
Presence of resistant infectious organisms
 MRSA 54.74 15.46, 94.03 < 0.001
 Clostridioides difficile −19.86 −94.4, 54.67 0.492
 ESBL 106.52 − 101.52, 314.56 0.187
In-dwelling catheter 45.83 6.72, 84.94 0.003

GEE is generalised estimating equations. DOT is days of therapy. CI is confidence interval. MRSA is methicillin-resistant Staphylococcus aureus. ESBL is extended spectrum beta-lactamases. Model includes fixed effects for facilities to adjust for clustering within facilities. Marginal effects can be interpreted as the change in DOT/1000 resident days associated with a change in a given independent variable from the base/reference level, independent of all the other covariates in the model