Table 3.

Multi-level multiple logistic regression of rubber dam use with restoration- and patient-level variables modeled simultaneously, by restoration-level and patient-level characteristics

Explanatory covariate	Parameter Estimate	Standard Error	p-value
Intercept	−2.61	0.37	<0.001
Restoration-level
Tooth and arch type
Maxillary molar	−0.18	0.40	ns
Maxillary premolar	−0.03	0.39	ns
Maxillary anterior	0.66	0.30	0.029
Mandibular molar	0.25	0.41	ns
Mandibular premolar	0.08	0.39	ns
Mandibular anterior [reference group]
Restoration classification
Class I	0.77	0.17	<0.001
Class II	1.42	0.19	<0.001
Class III	0.49	0.30	ns
Class IV	0.92	0.35	0.008
Class V [reference group]
Reason that the restoration was placed
Primary caries	0.57	0.19	0.002
Non-carious defect [reference group]
Patient-level
Patient’s ethnicity
Hispanic or Latino	0.42	0.15	0.006
Not Hispanic or Latino [reference group]
Patient has dental insurance or third party coverage
Yes	0.37	0.15	0.015
No [reference group]
Patient’s age
Less than 18 years old	0.92	0.18	<0.001
18–45 years old	0.95	0.16	<0.001
45–64 years old	0.73	0.21	0.004
65 years old or older [reference group]

The outcome of interest (use of a rubber dam) was coded 1 if a rubber dam was used on the restoration/patient, and 0 if not. n = 3,597; model fit: QIC = 4,184. This logistic regression was limited to the 3,714 restorations that were placed for the 2,109 patients by the 85 dentists who placed a rubber dam on at least one restoration at any time during the study. This was done with the GENMOD procedure in SAS, using a logit link, binomial distribution, exchangeable correlation structure. Although the restoration is the unit of analysis, clustering of restorations within patients and dentists is accounted for in the correlation structure, allowing for both restoration-specific and patient-specific characteristics to be tested.