Evaluation of a priority setting tool for triaging dental care
Objectives: To evaluate the capacity of 8 screening questions to validly identify priorities for dental care.
Method: Using a cross sectional study design, a convenience sample of 310 patients seeking dental care was selected from the first five callers per day requesting symptomatic care at 2 South Australian Dental Service (SADS) community dental clinics. Callers answered 8 questions concerning dental symptoms. Based on the same phone call, receptionists recorded their judgements of priority into three categories of urgency: <48 hours, 2-7 days, or 8+ days. Patients' "gold standard" priority for care was assessed by SADS dentists using the same three categories of urgency. Coefficients from a previously developed binary regression model predicting urgency of care in a separate sample of patients (N=784) were used to generate each patient's predicted probability of requiring care. Sensitivity(Se), specificity(Sp) and predictive values were computed against the dentist's gold standard priority for care that was dichotomised at 2 thresholds, <48 hours vs 2+days and; < 7 days vs 8+ days.
Results: At <48hour threshold, the model performed better than receptionists in predicting patients who did not need care (Se 56 % & Sp 70%, cf Se 62% & Sp 60%). However, at <7days threshold the model performed better than receptionists in predicting those who needed care (Se 80% & Sp 42% cf Se 56% & Sp 55%). Overall agreement between predicted urgency and dentists' gold standard was greater for the model (weighted Kappa = 0.25, 95% CI = 0.16-0.35) than for receptionists (weighted Kappa = 0.11, 95% CI=0.02 - 0.20).
Conclusion: Screening for dental care priority was more accurate using a questionnaire than receptionists' subjective assessment. However, false-positive and false-negative rates varied at different thresholds of priority and need to be valued explicitly in relation to public health priorities before adopting either screening method.
K Jones*, GD Slade, AJ Spencer
Presented at the 84th General Session and Exhibition of the IADR, 28 June - 1 July 2006, Brisbane Australia
Note: * indicates presenter
Method: Using a cross sectional study design, a convenience sample of 310 patients seeking dental care was selected from the first five callers per day requesting symptomatic care at 2 South Australian Dental Service (SADS) community dental clinics. Callers answered 8 questions concerning dental symptoms. Based on the same phone call, receptionists recorded their judgements of priority into three categories of urgency: <48 hours, 2-7 days, or 8+ days. Patients' "gold standard" priority for care was assessed by SADS dentists using the same three categories of urgency. Coefficients from a previously developed binary regression model predicting urgency of care in a separate sample of patients (N=784) were used to generate each patient's predicted probability of requiring care. Sensitivity(Se), specificity(Sp) and predictive values were computed against the dentist's gold standard priority for care that was dichotomised at 2 thresholds, <48 hours vs 2+days and; < 7 days vs 8+ days.
Results: At <48hour threshold, the model performed better than receptionists in predicting patients who did not need care (Se 56 % & Sp 70%, cf Se 62% & Sp 60%). However, at <7days threshold the model performed better than receptionists in predicting those who needed care (Se 80% & Sp 42% cf Se 56% & Sp 55%). Overall agreement between predicted urgency and dentists' gold standard was greater for the model (weighted Kappa = 0.25, 95% CI = 0.16-0.35) than for receptionists (weighted Kappa = 0.11, 95% CI=0.02 - 0.20).
Conclusion: Screening for dental care priority was more accurate using a questionnaire than receptionists' subjective assessment. However, false-positive and false-negative rates varied at different thresholds of priority and need to be valued explicitly in relation to public health priorities before adopting either screening method.
K Jones*, GD Slade, AJ Spencer
Presented at the 84th General Session and Exhibition of the IADR, 28 June - 1 July 2006, Brisbane Australia
Note: * indicates presenter