A Tree Model Analysis Exploring How Primary Care Doctors Diagnose Depression

Abstract

Aims: To explore the criteria used by primary care physicians (PCPs) to diagnose depression and the factors influencing their decisions using an inference decision tree analysis to how the factors interact. Methodology: 10,179 adult patients were recruited from the waiting rooms of 59 PCPs in Hong Kong. Patients completed a survey collecting data on socio-demographics and the Patient Health Questionnaire-9 (PHQ-9). Blinded doctors documented whether they thought the patient had depression (yes/no). Data was analyzed using multiple logistic regression and modelled to generate a conditional inference decision tree to reveal how PHQ-9 symptoms and PHQ-9 total scores were associated with diagnosis. Results: 1,054 patients received a depression diagnosis. Logistic regression identified gender, age, employment status, past history of depression, family history of mental illness and recent doctor visit as factors associated with diagnosis. Tree analyses revealed differing diagnostic pathways for patients with and without past depression. The PHQ-9 symptom model revealed low mood, sense of worthlessness, fatigue, sleep disturbance and functional impairment as early classifiers. The PHQ-9 score model revealed cut-off scores of >12 and >15 were used to diagnose patients with and without past depression Conclusion: The tree models demonstrated how PCPs use a hypothetical-deductive problem-solving approach incorporating pre-test probability, using different criteria for patients with and without past depression. Diagnostic thresholds may be too low for patients with past depression and too high for patients with no past depression potentially leading to over and under diagnosis of depression. Suicide assessment and prevention may be suboptimal in our setting.